Overview
Submit package home:enzokiel:branches:Essentials / x265 to package Essentials / x265
x265.changes
Changed
x
1
2
-------------------------------------------------------------------3
+Fri Dec 01 16:40:13 UTC 2017 - joerg.lorenzen@ki.tng.de4
+5
+- Update to version 2.66
+ New features7
+ * x265 can now refine analysis from a previous HEVC encode (using8
+ options --refine-inter, and --refine-intra), or a previous AVC9
+ encode (using option --refine-mv-type). The previous encode’s10
+ information can be packaged using the x265_analysis_data_t data11
+ field available in the x265_picture object.12
+ * Basic support for segmented (or chunked) encoding added with13
+ --vbv-end that can specify the status of CPB at the end of a14
+ segment. String this together with --vbv-init to encode a title15
+ as chunks while maintaining VBV compliance!16
+ * --force-flush can be used to trigger a premature flush of the17
+ encoder. This option is beneficial when input is known to be18
+ bursty, and may be at a rate slower than the encoder.19
+ * Experimental feature --lowpass-dct that uses truncated DCT for20
+ transformation.21
+ Encoder enhancements22
+ * Slice-parallel mode gets a significant boost in performance,23
+ particularly in low-latency mode.24
+ * x265 now officially supported on VS2017.25
+ * x265 now supports all depths from mono0 to mono16 for Y4M26
+ format.27
+ API changes28
+ * Options that modified PPS dynamically (--opt-qp-pps and29
+ --opt-ref-list-length-pps) are now disabled by default to30
+ enable users to save bits by not sending headers. If these31
+ options are enabled, headers have to be repeated for every GOP.32
+ * Rate-control and analysis parameters can dynamically be33
+ reconfigured simultaneously via the x265_encoder_reconfig API.34
+ * New API functions to extract intermediate information such as35
+ slice-type, scenecut information, reference frames, etc. are36
+ now available. This information may be beneficial to37
+ integrating applications that are attempting to perform38
+ content-adaptive encoding. Refer to documentation on39
+ x265_get_slicetype_poc_and_scenecut, and40
+ x265_get_ref_frame_list for more details and suggested usage.41
+ * A new API to pass supplemental CTU information to x265 to42
+ influence analysis decisions has been added. Refer to43
+ documentation on x265_encoder_ctu_info for more details.44
+ Bug fixes45
+ * Bug fixes when --slices is used with VBV settings.46
+ * Minor memory leak fixed for HDR10+ builds, and default x26547
+ when pools option is specified.48
+ * HDR10+ bug fix to remove dependence on poc counter to select49
+ meta-data information.50
+51
+-------------------------------------------------------------------52
Thu Jul 27 08:33:52 UTC 2017 - joerg.lorenzen@ki.tng.de53
54
- Update to version 2.555
x265.spec
Changed
23
1
2
# based on the spec file from https://build.opensuse.org/package/view_file/home:Simmphonie/libx265/3
4
Name: x2655
-%define soname 1306
+%define soname 1467
%define libname lib%{name}8
%define libsoname %{libname}-%{soname}9
-Version: 2.510
+Version: 2.611
Release: 012
License: GPL-2.0+13
Summary: A free h265/HEVC encoder - encoder binary14
15
streams. 16
17
%prep18
-%setup -q -n %{name}_%{version}19
+%setup -q -n %{name}_v%{version}20
%patch0 -p121
%patch1 -p122
23
x265_2.5.tar.gz/source/x265-extras.cpp
Deleted
449
1
2
-/*****************************************************************************3
- * Copyright (C) 2013-2017 MulticoreWare, Inc4
- *5
- * Authors: Steve Borho <steve@borho.org>6
- * Selvakumar Nithiyaruban <selvakumar@multicorewareinc.com>7
- * Divya Manivannan <divya@multicorewareinc.com>8
- *9
- * This program is free software; you can redistribute it and/or modify10
- * it under the terms of the GNU General Public License as published by11
- * the Free Software Foundation; either version 2 of the License, or12
- * (at your option) any later version.13
- *14
- * This program is distributed in the hope that it will be useful,15
- * but WITHOUT ANY WARRANTY; without even the implied warranty of16
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the17
- * GNU General Public License for more details.18
- *19
- * You should have received a copy of the GNU General Public License20
- * along with this program; if not, write to the Free Software21
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.22
- *23
- * This program is also available under a commercial proprietary license.24
- * For more information, contact us at license @ x265.com.25
- *****************************************************************************/26
-27
-#include "x265.h"28
-#include "x265-extras.h"29
-#include "param.h"30
-#include "common.h"31
-32
-using namespace X265_NS;33
-34
-static const char* summaryCSVHeader =35
- "Command, Date/Time, Elapsed Time, FPS, Bitrate, "36
- "Y PSNR, U PSNR, V PSNR, Global PSNR, SSIM, SSIM (dB), "37
- "I count, I ave-QP, I kbps, I-PSNR Y, I-PSNR U, I-PSNR V, I-SSIM (dB), "38
- "P count, P ave-QP, P kbps, P-PSNR Y, P-PSNR U, P-PSNR V, P-SSIM (dB), "39
- "B count, B ave-QP, B kbps, B-PSNR Y, B-PSNR U, B-PSNR V, B-SSIM (dB), "40
- "MaxCLL, MaxFALL, Version\n";41
-42
-FILE* x265_csvlog_open(const x265_param& param, const char* fname, int level)43
-{44
- FILE *csvfp = x265_fopen(fname, "r");45
- if (csvfp)46
- {47
- /* file already exists, re-open for append */48
- fclose(csvfp);49
- return x265_fopen(fname, "ab");50
- }51
- else52
- {53
- /* new CSV file, write header */54
- csvfp = x265_fopen(fname, "wb");55
- if (csvfp)56
- {57
- if (level)58
- {59
- fprintf(csvfp, "Encode Order, Type, POC, QP, Bits, Scenecut, ");60
- if (level >= 2)61
- fprintf(csvfp, "I/P cost ratio, ");62
- if (param.rc.rateControlMode == X265_RC_CRF)63
- fprintf(csvfp, "RateFactor, ");64
- if (param.rc.vbvBufferSize)65
- fprintf(csvfp, "BufferFill, ");66
- if (param.bEnablePsnr)67
- fprintf(csvfp, "Y PSNR, U PSNR, V PSNR, YUV PSNR, ");68
- if (param.bEnableSsim)69
- fprintf(csvfp, "SSIM, SSIM(dB), ");70
- fprintf(csvfp, "Latency, ");71
- fprintf(csvfp, "List 0, List 1");72
- uint32_t size = param.maxCUSize;73
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)74
- {75
- fprintf(csvfp, ", Intra %dx%d DC, Intra %dx%d Planar, Intra %dx%d Ang", size, size, size, size, size, size);76
- size /= 2;77
- }78
- fprintf(csvfp, ", 4x4");79
- size = param.maxCUSize;80
- if (param.bEnableRectInter)81
- {82
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)83
- {84
- fprintf(csvfp, ", Inter %dx%d, Inter %dx%d (Rect)", size, size, size, size);85
- if (param.bEnableAMP)86
- fprintf(csvfp, ", Inter %dx%d (Amp)", size, size);87
- size /= 2;88
- }89
- }90
- else91
- {92
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)93
- {94
- fprintf(csvfp, ", Inter %dx%d", size, size);95
- size /= 2;96
- }97
- }98
- size = param.maxCUSize;99
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)100
- {101
- fprintf(csvfp, ", Skip %dx%d", size, size);102
- size /= 2;103
- }104
- size = param.maxCUSize;105
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)106
- {107
- fprintf(csvfp, ", Merge %dx%d", size, size);108
- size /= 2;109
- }110
-111
- if (level >= 2)112
- {113
- fprintf(csvfp, ", Avg Luma Distortion, Avg Chroma Distortion, Avg psyEnergy, Avg Residual Energy,"114
- " Min Luma Level, Max Luma Level, Avg Luma Level");115
-116
- if (param.internalCsp != X265_CSP_I400)117
- fprintf(csvfp, ", Min Cb Level, Max Cb Level, Avg Cb Level, Min Cr Level, Max Cr Level, Avg Cr Level");118
-119
- /* PU statistics */120
- size = param.maxCUSize;121
- for (uint32_t i = 0; i< param.maxLog2CUSize - (uint32_t)g_log2Size[param.minCUSize] + 1; i++)122
- {123
- fprintf(csvfp, ", Intra %dx%d", size, size);124
- fprintf(csvfp, ", Skip %dx%d", size, size);125
- fprintf(csvfp, ", AMP %d", size);126
- fprintf(csvfp, ", Inter %dx%d", size, size);127
- fprintf(csvfp, ", Merge %dx%d", size, size);128
- fprintf(csvfp, ", Inter %dx%d", size, size / 2);129
- fprintf(csvfp, ", Merge %dx%d", size, size / 2);130
- fprintf(csvfp, ", Inter %dx%d", size / 2, size);131
- fprintf(csvfp, ", Merge %dx%d", size / 2, size);132
- size /= 2;133
- }134
-135
- if ((uint32_t)g_log2Size[param.minCUSize] == 3)136
- fprintf(csvfp, ", 4x4");137
-138
- /* detailed performance statistics */139
- fprintf(csvfp, ", DecideWait (ms), Row0Wait (ms), Wall time (ms), Ref Wait Wall (ms), Total CTU time (ms),"140
- "Stall Time (ms), Total frame time (ms), Avg WPP, Row Blocks");141
- }142
- fprintf(csvfp, "\n");143
- }144
- else145
- fputs(summaryCSVHeader, csvfp);146
- }147
- return csvfp;148
- }149
-}150
-151
-// per frame CSV logging152
-void x265_csvlog_frame(FILE* csvfp, const x265_param& param, const x265_picture& pic, int level)153
-{154
- if (!csvfp)155
- return;156
-157
- const x265_frame_stats* frameStats = &pic.frameData;158
- fprintf(csvfp, "%d, %c-SLICE, %4d, %2.2lf, %10d, %d,", frameStats->encoderOrder, frameStats->sliceType, frameStats->poc, 159
- frameStats->qp, (int)frameStats->bits, frameStats->bScenecut);160
- if (level >= 2)161
- fprintf(csvfp, "%.2f,", frameStats->ipCostRatio);162
- if (param.rc.rateControlMode == X265_RC_CRF)163
- fprintf(csvfp, "%.3lf,", frameStats->rateFactor);164
- if (param.rc.vbvBufferSize)165
- fprintf(csvfp, "%.3lf,", frameStats->bufferFill);166
- if (param.bEnablePsnr)167
- fprintf(csvfp, "%.3lf, %.3lf, %.3lf, %.3lf,", frameStats->psnrY, frameStats->psnrU, frameStats->psnrV, frameStats->psnr);168
- if (param.bEnableSsim)169
- fprintf(csvfp, " %.6f, %6.3f,", frameStats->ssim, x265_ssim2dB(frameStats->ssim));170
- fprintf(csvfp, "%d, ", frameStats->frameLatency);171
- if (frameStats->sliceType == 'I' || frameStats->sliceType == 'i')172
- fputs(" -, -,", csvfp);173
- else174
- {175
- int i = 0;176
- while (frameStats->list0POC[i] != -1)177
- fprintf(csvfp, "%d ", frameStats->list0POC[i++]);178
- fprintf(csvfp, ",");179
- if (frameStats->sliceType != 'P')180
- {181
- i = 0;182
- while (frameStats->list1POC[i] != -1)183
- fprintf(csvfp, "%d ", frameStats->list1POC[i++]);184
- fprintf(csvfp, ",");185
- }186
- else187
- fputs(" -,", csvfp);188
- }189
-190
- if (level)191
- {192
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)193
- fprintf(csvfp, "%5.2lf%%, %5.2lf%%, %5.2lf%%,", frameStats->cuStats.percentIntraDistribution[depth][0],194
- frameStats->cuStats.percentIntraDistribution[depth][1],195
- frameStats->cuStats.percentIntraDistribution[depth][2]);196
- fprintf(csvfp, "%5.2lf%%", frameStats->cuStats.percentIntraNxN);197
- if (param.bEnableRectInter)198
- {199
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)200
- {201
- fprintf(csvfp, ", %5.2lf%%, %5.2lf%%", frameStats->cuStats.percentInterDistribution[depth][0],202
- frameStats->cuStats.percentInterDistribution[depth][1]);203
- if (param.bEnableAMP)204
- fprintf(csvfp, ", %5.2lf%%", frameStats->cuStats.percentInterDistribution[depth][2]);205
- }206
- }207
- else208
- {209
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)210
- fprintf(csvfp, ", %5.2lf%%", frameStats->cuStats.percentInterDistribution[depth][0]);211
- }212
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)213
- fprintf(csvfp, ", %5.2lf%%", frameStats->cuStats.percentSkipCu[depth]);214
- for (uint32_t depth = 0; depth <= param.maxCUDepth; depth++)215
- fprintf(csvfp, ", %5.2lf%%", frameStats->cuStats.percentMergeCu[depth]);216
- }217
-218
- if (level >= 2)219
- {220
- fprintf(csvfp, ", %.2lf, %.2lf, %.2lf, %.2lf ", frameStats->avgLumaDistortion,221
- frameStats->avgChromaDistortion,222
- frameStats->avgPsyEnergy,223
- frameStats->avgResEnergy);224
-225
- fprintf(csvfp, ", %d, %d, %.2lf", frameStats->minLumaLevel, frameStats->maxLumaLevel, frameStats->avgLumaLevel);226
-227
- if (param.internalCsp != X265_CSP_I400)228
- {229
- fprintf(csvfp, ", %d, %d, %.2lf", frameStats->minChromaULevel, frameStats->maxChromaULevel, frameStats->avgChromaULevel);230
- fprintf(csvfp, ", %d, %d, %.2lf", frameStats->minChromaVLevel, frameStats->maxChromaVLevel, frameStats->avgChromaVLevel);231
- }232
-233
- for (uint32_t i = 0; i < param.maxLog2CUSize - (uint32_t)g_log2Size[param.minCUSize] + 1; i++)234
- {235
- fprintf(csvfp, ", %.2lf%%", frameStats->puStats.percentIntraPu[i]);236
- fprintf(csvfp, ", %.2lf%%", frameStats->puStats.percentSkipPu[i]);237
- fprintf(csvfp, ",%.2lf%%", frameStats->puStats.percentAmpPu[i]);238
- for (uint32_t j = 0; j < 3; j++)239
- {240
- fprintf(csvfp, ", %.2lf%%", frameStats->puStats.percentInterPu[i][j]);241
- fprintf(csvfp, ", %.2lf%%", frameStats->puStats.percentMergePu[i][j]);242
- }243
- }244
- if ((uint32_t)g_log2Size[param.minCUSize] == 3)245
- fprintf(csvfp, ",%.2lf%%", frameStats->puStats.percentNxN);246
-247
- fprintf(csvfp, ", %.1lf, %.1lf, %.1lf, %.1lf, %.1lf, %.1lf, %.1lf,", frameStats->decideWaitTime, frameStats->row0WaitTime,248
- frameStats->wallTime, frameStats->refWaitWallTime,249
- frameStats->totalCTUTime, frameStats->stallTime,250
- frameStats->totalFrameTime);251
-252
- fprintf(csvfp, " %.3lf, %d", frameStats->avgWPP, frameStats->countRowBlocks);253
- }254
- fprintf(csvfp, "\n");255
- fflush(stderr);256
-}257
-258
-void x265_csvlog_encode(FILE* csvfp, const char* version, const x265_param& param, int padx, int pady, const x265_stats& stats, int level, int argc, char** argv)259
-{260
- if (!csvfp)261
- return;262
-263
- if (level)264
- {265
- // adding summary to a per-frame csv log file, so it needs a summary header266
- fprintf(csvfp, "\nSummary\n");267
- fputs(summaryCSVHeader, csvfp);268
- }269
-270
- // CLI arguments or other271
- if (argc)272
- {273
- fputc('"', csvfp);274
- for (int i = 1; i < argc; i++)275
- {276
- fputc(' ', csvfp);277
- fputs(argv[i], csvfp);278
- }279
- fputc('"', csvfp);280
- }281
- else282
- {283
- const x265_param* paramTemp = ¶m;284
- char *opts = x265_param2string((x265_param*)paramTemp, padx, pady);285
- if (opts)286
- {287
- fputc('"', csvfp);288
- fputs(opts, csvfp);289
- fputc('"', csvfp);290
- }291
- }292
-293
- // current date and time294
- time_t now;295
- struct tm* timeinfo;296
- time(&now);297
- timeinfo = localtime(&now);298
- char buffer[200];299
- strftime(buffer, 128, "%c", timeinfo);300
- fprintf(csvfp, ", %s, ", buffer);301
-302
- // elapsed time, fps, bitrate303
- fprintf(csvfp, "%.2f, %.2f, %.2f,",304
- stats.elapsedEncodeTime, stats.encodedPictureCount / stats.elapsedEncodeTime, stats.bitrate);305
-306
- if (param.bEnablePsnr)307
- fprintf(csvfp, " %.3lf, %.3lf, %.3lf, %.3lf,",308
- stats.globalPsnrY / stats.encodedPictureCount, stats.globalPsnrU / stats.encodedPictureCount,309
- stats.globalPsnrV / stats.encodedPictureCount, stats.globalPsnr);310
- else311
- fprintf(csvfp, " -, -, -, -,");312
- if (param.bEnableSsim)313
- fprintf(csvfp, " %.6f, %6.3f,", stats.globalSsim, x265_ssim2dB(stats.globalSsim));314
- else315
- fprintf(csvfp, " -, -,");316
-317
- if (stats.statsI.numPics)318
- {319
- fprintf(csvfp, " %-6u, %2.2lf, %-8.2lf,", stats.statsI.numPics, stats.statsI.avgQp, stats.statsI.bitrate);320
- if (param.bEnablePsnr)321
- fprintf(csvfp, " %.3lf, %.3lf, %.3lf,", stats.statsI.psnrY, stats.statsI.psnrU, stats.statsI.psnrV);322
- else323
- fprintf(csvfp, " -, -, -,");324
- if (param.bEnableSsim)325
- fprintf(csvfp, " %.3lf,", stats.statsI.ssim);326
- else327
- fprintf(csvfp, " -,");328
- }329
- else330
- fprintf(csvfp, " -, -, -, -, -, -, -,");331
-332
- if (stats.statsP.numPics)333
- {334
- fprintf(csvfp, " %-6u, %2.2lf, %-8.2lf,", stats.statsP.numPics, stats.statsP.avgQp, stats.statsP.bitrate);335
- if (param.bEnablePsnr)336
- fprintf(csvfp, " %.3lf, %.3lf, %.3lf,", stats.statsP.psnrY, stats.statsP.psnrU, stats.statsP.psnrV);337
- else338
- fprintf(csvfp, " -, -, -,");339
- if (param.bEnableSsim)340
- fprintf(csvfp, " %.3lf,", stats.statsP.ssim);341
- else342
- fprintf(csvfp, " -,");343
- }344
- else345
- fprintf(csvfp, " -, -, -, -, -, -, -,");346
-347
- if (stats.statsB.numPics)348
- {349
- fprintf(csvfp, " %-6u, %2.2lf, %-8.2lf,", stats.statsB.numPics, stats.statsB.avgQp, stats.statsB.bitrate);350
- if (param.bEnablePsnr)351
- fprintf(csvfp, " %.3lf, %.3lf, %.3lf,", stats.statsB.psnrY, stats.statsB.psnrU, stats.statsB.psnrV);352
- else353
- fprintf(csvfp, " -, -, -,");354
- if (param.bEnableSsim)355
- fprintf(csvfp, " %.3lf,", stats.statsB.ssim);356
- else357
- fprintf(csvfp, " -,");358
- }359
- else360
- fprintf(csvfp, " -, -, -, -, -, -, -,");361
-362
- fprintf(csvfp, " %-6u, %-6u, %s\n", stats.maxCLL, stats.maxFALL, version);363
-}364
-365
-/* The dithering algorithm is based on Sierra-2-4A error diffusion.366
- * We convert planes in place (without allocating a new buffer). */367
-static void ditherPlane(uint16_t *src, int srcStride, int width, int height, int16_t *errors, int bitDepth)368
-{369
- const int lShift = 16 - bitDepth;370
- const int rShift = 16 - bitDepth + 2;371
- const int half = (1 << (16 - bitDepth + 1));372
- const int pixelMax = (1 << bitDepth) - 1;373
-374
- memset(errors, 0, (width + 1) * sizeof(int16_t));375
-376
- if (bitDepth == 8)377
- {378
- for (int y = 0; y < height; y++, src += srcStride)379
- {380
- uint8_t* dst = (uint8_t *)src;381
- int16_t err = 0;382
- for (int x = 0; x < width; x++)383
- {384
- err = err * 2 + errors[x] + errors[x + 1];385
- int tmpDst = x265_clip3(0, pixelMax, ((src[x] << 2) + err + half) >> rShift);386
- errors[x] = err = (int16_t)(src[x] - (tmpDst << lShift));387
- dst[x] = (uint8_t)tmpDst;388
- }389
- }390
- }391
- else392
- {393
- for (int y = 0; y < height; y++, src += srcStride)394
- {395
- int16_t err = 0;396
- for (int x = 0; x < width; x++)397
- {398
- err = err * 2 + errors[x] + errors[x + 1];399
- int tmpDst = x265_clip3(0, pixelMax, ((src[x] << 2) + err + half) >> rShift);400
- errors[x] = err = (int16_t)(src[x] - (tmpDst << lShift));401
- src[x] = (uint16_t)tmpDst;402
- }403
- }404
- }405
-}406
-407
-void x265_dither_image(const x265_api& api, x265_picture& picIn, int picWidth, int picHeight, int16_t *errorBuf, int bitDepth)408
-{409
- if (sizeof(x265_picture) != api.sizeof_picture)410
- {411
- fprintf(stderr, "extras [error]: structure size skew, unable to dither\n");412
- return;413
- }414
-415
- if (picIn.bitDepth <= 8)416
- {417
- fprintf(stderr, "extras [error]: dither support enabled only for input bitdepth > 8\n");418
- return;419
- }420
-421
- if (picIn.bitDepth == bitDepth)422
- {423
- fprintf(stderr, "extras[error]: dither support enabled only if encoder depth is different from picture depth\n");424
- return;425
- }426
-427
- /* This portion of code is from readFrame in x264. */428
- for (int i = 0; i < x265_cli_csps[picIn.colorSpace].planes; i++)429
- {430
- if (picIn.bitDepth < 16)431
- {432
- /* upconvert non 16bit high depth planes to 16bit */433
- uint16_t *plane = (uint16_t*)picIn.planes[i];434
- uint32_t pixelCount = x265_picturePlaneSize(picIn.colorSpace, picWidth, picHeight, i);435
- int lShift = 16 - picIn.bitDepth;436
-437
- /* This loop assumes width is equal to stride which438
- * happens to be true for file reader outputs */439
- for (uint32_t j = 0; j < pixelCount; j++)440
- plane[j] = plane[j] << lShift;441
- }442
-443
- int height = (int)(picHeight >> x265_cli_csps[picIn.colorSpace].height[i]);444
- int width = (int)(picWidth >> x265_cli_csps[picIn.colorSpace].width[i]);445
-446
- ditherPlane(((uint16_t*)picIn.planes[i]), picIn.stride[i] / 2, width, height, errorBuf, bitDepth);447
- }448
-}449
x265_2.5.tar.gz/source/x265-extras.h
Deleted
68
1
2
-/*****************************************************************************3
- * Copyright (C) 2013-2017 MulticoreWare, Inc4
- *5
- * Authors: Steve Borho <steve@borho.org>6
- *7
- * This program is free software; you can redistribute it and/or modify8
- * it under the terms of the GNU General Public License as published by9
- * the Free Software Foundation; either version 2 of the License, or10
- * (at your option) any later version.11
- *12
- * This program is distributed in the hope that it will be useful,13
- * but WITHOUT ANY WARRANTY; without even the implied warranty of14
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the15
- * GNU General Public License for more details.16
- *17
- * You should have received a copy of the GNU General Public License18
- * along with this program; if not, write to the Free Software19
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.20
- *21
- * This program is also available under a commercial proprietary license.22
- * For more information, contact us at license @ x265.com.23
- *****************************************************************************/24
-25
-#ifndef X265_EXTRAS_H26
-#define X265_EXTRAS_H 127
-28
-#include "x265.h"29
-30
-#include <stdio.h>31
-#include <stdint.h>32
-33
-#ifdef __cplusplus34
-extern "C" {35
-#endif36
-37
-#if _WIN3238
-#define LIBAPI __declspec(dllexport)39
-#else40
-#define LIBAPI41
-#endif42
-43
-/* Open a CSV log file. On success it returns a file handle which must be passed44
- * to x265_csvlog_frame() and/or x265_csvlog_encode(). The file handle must be45
- * closed by the caller using fclose(). If level is 0, then no frame logging46
- * header is written to the file. This function will return NULL if it is unable47
- * to open the file for write or if it detects a structure size skew */48
-LIBAPI FILE* x265_csvlog_open(const x265_param& param, const char* fname, int level);49
-50
-/* Log frame statistics to the CSV file handle. level should have been non-zero51
- * in the call to x265_csvlog_open() if this function is called. */52
-LIBAPI void x265_csvlog_frame(FILE* csvfp, const x265_param& param, const x265_picture& pic, int level);53
-54
-/* Log final encode statistics to the CSV file handle. 'argc' and 'argv' are55
- * intended to be command line arguments passed to the encoder. Encode56
- * statistics should be queried from the encoder just prior to closing it. */57
-LIBAPI void x265_csvlog_encode(FILE* csvfp, const char* version, const x265_param& param, int padx, int pady, const x265_stats& stats, int level, int argc, char** argv);58
-59
-/* In-place downshift from a bit-depth greater than 8 to a bit-depth of 8, using60
- * the residual bits to dither each row. */61
-LIBAPI void x265_dither_image(const x265_api& api, x265_picture&, int picWidth, int picHeight, int16_t *errorBuf, int bitDepth);62
-63
-#ifdef __cplusplus64
-}65
-#endif66
-67
-#endif68
x265_2.5.tar.gz/.hg_archival.txt -> x265_2.6.tar.gz/.hg_archival.txt
Changed
8
1
2
repo: 09fe40627f03a0f9c3e6ac78b22ac93da23f9fdf3
-node: 64b2d0bf45a52511e57a6b7299160b961ca3d51c4
+node: 0e9ea76945c89962cd46cee6537586e2054b29355
branch: stable6
-tag: 2.57
+tag: 2.68
x265_2.5.tar.gz/.hgtags -> x265_2.6.tar.gz/.hgtags
Changed
6
1
2
be14a7e9755e54f0fd34911c72bdfa66981220bc 2.23
3037c1448549ca920967831482c653e5892fa8ed 2.34
e7a4dd48293b7956d4a20df257d23904cc78e376 2.45
+64b2d0bf45a52511e57a6b7299160b961ca3d51c 2.56
x265_2.6.tar.gz/build/msys-cl/make-Makefiles-64bit.sh
Added
30
1
2
+#!/bin/sh3
+# This is to generate visual studio builds with required environment variables set in this shell, useful for ffmpeg integration4
+# Run this from within an MSYS bash shell5
+6
+target_processor='amd64'7
+path=$(which cl)8
+9
+if cl; then10
+ echo11
+else12
+ echo "please launch 'visual studio command prompt' and run '..\vcvarsall.bat amd64'"13
+ echo "and then launch msys bash shell from there"14
+ exit 115
+fi16
+17
+if [[ $path == *$target_processor* ]]; then18
+ echo19
+else20
+ echo "64 bit target not set, please launch 'visual studio command prompt' and run '..\vcvarsall.bat amd64 | x86_amd64 | amd64_x86'"21
+ exit 122
+fi23
+24
+cmake -G "NMake Makefiles" -DCMAKE_CXX_FLAGS="-DWIN32 -D_WINDOWS -W4 -GR -EHsc" -DCMAKE_C_FLAGS="-DWIN32 -D_WINDOWS -W4" ../../source25
+if [ -e Makefile ]26
+then27
+ nmake28
+fi29
\ No newline at end of file30
x265_2.6.tar.gz/build/msys-cl/make-Makefiles.sh
Added
20
1
2
+#!/bin/sh3
+# This is to generate visual studio builds with required environment variables set in this shell, useful for ffmpeg integration4
+# Run this from within an MSYS bash shell5
+6
+if cl; then7
+ echo 8
+else9
+ echo "please launch msys from 'visual studio command prompt'"10
+ exit 111
+fi12
+13
+cmake -G "NMake Makefiles" -DCMAKE_CXX_FLAGS="-DWIN32 -D_WINDOWS -W4 -GR -EHsc" -DCMAKE_C_FLAGS="-DWIN32 -D_WINDOWS -W4" ../../source14
+15
+if [ -e Makefile ]16
+then17
+ nmake18
+fi19
\ No newline at end of file20
x265_2.6.tar.gz/build/vc15-x86/build-all.bat
Added
16
1
2
+@echo off3
+if "%VS150COMNTOOLS%" == "" (4
+ msg "%username%" "Visual Studio 15 not detected"5
+ exit 16
+)7
+if not exist x265.sln (8
+ call make-solutions.bat9
+)10
+if exist x265.sln (11
+ call "%VS150COMNTOOLS%\..\..\VC\vcvarsall.bat"12
+ MSBuild /property:Configuration="Release" x265.sln13
+ MSBuild /property:Configuration="Debug" x265.sln14
+ MSBuild /property:Configuration="RelWithDebInfo" x265.sln15
+)16
x265_2.6.tar.gz/build/vc15-x86/make-solutions.bat
Added
8
1
2
+@echo off3
+::4
+:: run this batch file to create a Visual Studio solution file for this project.5
+:: See the cmake documentation for other generator targets6
+::7
+cmake -G "Visual Studio 15" ..\..\source && cmake-gui ..\..\source8
x265_2.6.tar.gz/build/vc15-x86_64/build-all.bat
Added
16
1
2
+@echo off3
+if "%VS150COMNTOOLS%" == "" (4
+ msg "%username%" "Visual Studio 15 not detected"5
+ exit 16
+)7
+if not exist x265.sln (8
+ call make-solutions.bat9
+)10
+if exist x265.sln (11
+ call "%VS150COMNTOOLS%\..\..\VC\vcvarsall.bat"12
+ MSBuild /property:Configuration="Release" x265.sln13
+ MSBuild /property:Configuration="Debug" x265.sln14
+ MSBuild /property:Configuration="RelWithDebInfo" x265.sln15
+)16
x265_2.6.tar.gz/build/vc15-x86_64/make-solutions.bat
Added
8
1
2
+@echo off3
+::4
+:: run this batch file to create a Visual Studio solution file for this project.5
+:: See the cmake documentation for other generator targets6
+::7
+cmake -G "Visual Studio 15 Win64" ..\..\source && cmake-gui ..\..\source8
x265_2.6.tar.gz/build/vc15-x86_64/multilib.bat
Added
46
1
2
+@echo off3
+if "%VS150COMNTOOLS%" == "" (4
+ msg "%username%" "Visual Studio 15 not detected"5
+ exit 16
+)7
+8
+call "%VS150COMNTOOLS%\..\..\VC\vcvarsall.bat"9
+10
+@mkdir 12bit11
+@mkdir 10bit12
+@mkdir 8bit13
+14
+@cd 12bit15
+cmake -G "Visual Studio 15 Win64" ../../../source -DHIGH_BIT_DEPTH=ON -DEXPORT_C_API=OFF -DENABLE_SHARED=OFF -DENABLE_CLI=OFF -DMAIN12=ON16
+if exist x265.sln (17
+ MSBuild /property:Configuration="Release" x265.sln18
+ copy/y Release\x265-static.lib ..\8bit\x265-static-main12.lib19
+)20
+21
+@cd ..\10bit22
+cmake -G "Visual Studio 15 Win64" ../../../source -DHIGH_BIT_DEPTH=ON -DEXPORT_C_API=OFF -DENABLE_SHARED=OFF -DENABLE_CLI=OFF23
+if exist x265.sln (24
+ MSBuild /property:Configuration="Release" x265.sln25
+ copy/y Release\x265-static.lib ..\8bit\x265-static-main10.lib26
+)27
+28
+@cd ..\8bit29
+if not exist x265-static-main10.lib (30
+ msg "%username%" "10bit build failed"31
+ exit 132
+)33
+if not exist x265-static-main12.lib (34
+ msg "%username%" "12bit build failed"35
+ exit 136
+)37
+cmake -G "Visual Studio 15 Win64" ../../../source -DEXTRA_LIB="x265-static-main10.lib;x265-static-main12.lib" -DLINKED_10BIT=ON -DLINKED_12BIT=ON38
+if exist x265.sln (39
+ MSBuild /property:Configuration="Release" x265.sln40
+ :: combine static libraries (ignore warnings caused by winxp.cpp hacks)41
+ move Release\x265-static.lib x265-static-main.lib42
+ LIB.EXE /ignore:4006 /ignore:4221 /OUT:Release\x265-static.lib x265-static-main.lib x265-static-main10.lib x265-static-main12.lib43
+)44
+45
+pause46
x265_2.5.tar.gz/doc/reST/api.rst -> x265_2.6.tar.gz/doc/reST/api.rst
Changed
44
1
2
* presets is not recommended without a more fine-grained breakdown of3
* parameters to take this into account. */4
int x265_encoder_reconfig(x265_encoder *, x265_param *);5
-**x265_encoder_ctu_info**6
- /* x265_encoder_ctu_info:7
- * Copy CTU information such as ctu address and ctu partition structure of all8
- * CTUs in each frame. The function is invoked only if "--ctu-info" is enabled and9
- * the encoder will wait for this copy to complete if enabled.10
- */11
+12
+**x265_get_slicetype_poc_and_scenecut()** may be used to fetch slice type, poc and scene cut information mid-encode::13
+14
+ /* x265_get_slicetype_poc_and_scenecut:15
+ * get the slice type, poc and scene cut information for the current frame,16
+ * returns negative on error, 0 on success.17
+ * This API must be called after(poc >= lookaheadDepth + bframes + 2) condition check. */18
+ int x265_get_slicetype_poc_and_scenecut(x265_encoder *encoder, int *slicetype, int *poc, int* sceneCut);19
+20
+**x265_get_ref_frame_list()** may be used to fetch forward and backward refrence list::21
+22
+ /* x265_get_ref_frame_list:23
+ * returns negative on error, 0 when access unit were output.24
+ * This API must be called after(poc >= lookaheadDepth + bframes + 2) condition check */25
+ int x265_get_ref_frame_list(x265_encoder *encoder, x265_picyuv**, x265_picyuv**, int, int);26
+ 27
+**x265_encoder_ctu_info** may be used to provide additional CTU-specific information to the encoder::28
+29
+ /* x265_encoder_ctu_info:30
+ * Copy CTU information such as ctu address and ctu partition structure of all31
+ * CTUs in each frame. The function is invoked only if "--ctu-info" is enabled and32
+ * the encoder will wait for this copy to complete if enabled.*/33
+ int x265_encoder_ctu_info(x265_encoder *encoder, int poc, x265_ctu_info_t** ctu);34
+35
+**x265_set_analysis_data()** may be used to recive analysis information from external application::36
+37
+ /* x265_set_analysis_data:38
+ * set the analysis data. The incoming analysis_data structure is assumed to be AVC-sized blocks.39
+ * returns negative on error, 0 access unit were output.*/40
+ int x265_set_analysis_data(x265_encoder *encoder, x265_analysis_data *analysis_data, int poc, uint32_t cuBytes);41
42
Pictures43
========44
x265_2.5.tar.gz/doc/reST/cli.rst -> x265_2.6.tar.gz/doc/reST/cli.rst
Changed
296
1
2
3
Default: 1 slice per frame. **Experimental feature**4
5
+.. option:: --copy-pic, --no-copy-pic6
+7
+ Allow encoder to copy input x265 pictures to internal frame buffers. When disabled,8
+ x265 will not make an internal copy of the input picture and will work with the9
+ application's buffers. While this allows for deeper integration, it is the responsbility10
+ of the application to (a) ensure that the allocated picture has extra space for padding11
+ that will be done by the library, and (b) the buffers aren't recycled until the library12
+ has completed encoding this frame (which can be figured out by tracking NALs output by x265)13
+14
+ Default: enabled15
+16
+17
Input/Output File Options18
=========================19
20
21
22
Note that --analysis-reuse-level must be paired with analysis-reuse-mode.23
24
- +--------+-----------------------------------------+25
- | Level | Description |26
- +========+=========================================+27
- | 1 | Lookahead information |28
- +--------+-----------------------------------------+29
- | 2 to 4 | Level 1 + intra/inter modes, ref's |30
- +--------+-----------------------------------------+31
- | 5 to 9 | Level 2 + rect-amp |32
- +--------+-----------------------------------------+33
- | 10 | Level 5 + Full CU analysis-info |34
- +--------+-----------------------------------------+35
+ +--------------+------------------------------------------+36
+ | Level | Description |37
+ +==============+==========================================+38
+ | 1 | Lookahead information |39
+ +--------------+------------------------------------------+40
+ | 2 to 4 | Level 1 + intra/inter modes, ref's |41
+ +--------------+------------------------------------------+42
+ | 5,6 and 9 | Level 2 + rect-amp |43
+ +--------------+------------------------------------------+44
+ | 7 | Level 5 + AVC size CU refinement |45
+ +--------------+------------------------------------------+46
+ | 8 | Level 5 + AVC size Full CU analysis-info |47
+ +--------------+------------------------------------------+48
+ | 10 | Level 5 + Full CU analysis-info |49
+ +--------------+------------------------------------------+50
+51
+.. option:: --refine-mv-type <string>52
+53
+ Reuse MV information received through API call. Currently receives information for AVC size and the accepted 54
+ string input is "avc". Default is disabled.55
56
.. option:: --scale-factor57
58
59
This option should be coupled with analysis-reuse-mode option, --analysis-reuse-level 10.60
The ctu size of load should be double the size of save. Default 0.61
62
-.. option:: --refine-intra <0|1|2>63
+.. option:: --refine-intra <0..3>64
65
Enables refinement of intra blocks in current encode. 66
67
- Level 0 - Forces both mode and depth from the previous encode.68
+ Level 0 - Forces both mode and depth from the save encode.69
+ 70
+ Level 1 - Evaluates all intra modes at current depth(n) and at depth 71
+ (n+1) when current block size is one greater than the min-cu-size.72
+ Forces modes for larger blocks.73
74
- Level 1 - Evaluates all intra modes for blocks of size one smaller than 75
- the min-cu-size of the incoming analysis data from the previous encode, 76
- forces modes for blocks of larger size.77
+ Level 2 - In addition to the functionality of level 1, at all depths, force 78
+ (a) only depth when angular mode is chosen by the save encode.79
+ (b) depth and mode when other intra modes are chosen by the save encode.80
81
- Level 2 - Evaluates all intra modes for blocks of size one smaller than 82
- the min-cu-size of the incoming analysis data from the previous encode. 83
- For larger blocks, force only depth when angular mode is chosen by the 84
- previous encode, force depth and mode when other intra modes are chosen.85
+ Level 3 - Perform analysis of intra modes for depth reused from first encode.86
87
Default 0.88
89
-.. option:: --refine-inter-depth90
+.. option:: --refine-inter <0..3>91
92
- Enables refinement of inter blocks in current encode. Evaluates all 93
- inter modes for blocks of size one smaller than the min-cu-size of the 94
- incoming analysis data from the previous encode. Default disabled.95
+ Enables refinement of inter blocks in current encode. 96
+ 97
+ Level 0 - Forces both mode and depth from the save encode.98
+ 99
+ Level 1 - Evaluates all inter modes at current depth(n) and at depth 100
+ (n+1) when current block size is one greater than the min-cu-size.101
+ Forces modes for larger blocks.102
+ 103
+ Level 2 - In addition to the functionality of level 1, restricts the modes 104
+ evaluated when specific modes are decided as the best mode by the save encode.105
+ 106
+ 2nx2n in save encode - disable re-evaluation of rect and amp.107
+ 108
+ skip in save encode - re-evaluates only skip, merge and 2nx2n modes.109
+ 110
+ Level 3 - Perform analysis of inter modes while reusing depths from the save encode.111
+ 112
+ Default 0.113
114
.. option:: --refine-mv115
116
117
.. option:: --b-pyramid, --no-b-pyramid118
119
Use B-frames as references, when possible. Default enabled120
+ 121
+.. option:: --force-flush <integer>122
+123
+ Force the encoder to flush frames. Default is 0.124
+ 125
+ Values:126
+ 0 - flush the encoder only when all the input pictures are over.127
+ 1 - flush all the frames even when the input is not over. 128
+ slicetype decision may change with this option.129
+ 2 - flush the slicetype decided frames only. 130
131
Quality, rate control and rate distortion options132
=================================================133
134
Default 0.9135
136
**Range of values:** fractional: 0 - 1.0, or kbits: 2 .. bufsize137
+ 138
+.. option:: --vbv-end <float>139
+140
+ Final buffer emptiness. The portion of the decode buffer that must be 141
+ available after all the specified frames have been inserted into the 142
+ decode buffer. Specified as a fractional value between 0 and 1, or in 143
+ kbits. Default 0 (disabled)144
+ 145
+ This enables basic support for chunk-parallel encoding where each segment 146
+ can specify the starting and ending state of the VBV buffer so that VBV 147
+ compliance can be maintained when chunks are independently encoded and 148
+ stitched together.149
+ 150
+.. option:: --vbv-end-fr-adj <float>151
+152
+ Frame from which qp has to be adjusted to achieve final decode buffer153
+ emptiness. Specified as a fraction of the total frames. Fractions > 0 are 154
+ supported only when the total number of frames is known. Default 0.155
156
.. option:: --qp, -q <integer>157
158
159
Enable adaptive quantization for sub-CTUs. This parameter specifies 160
the minimum CU size at which QP can be adjusted, ie. Quantization Group161
size. Allowed range of values are 64, 32, 16, 8 provided this falls within 162
- the inclusive range [maxCUSize, minCUSize]. Experimental.163
+ the inclusive range [maxCUSize, minCUSize].164
Default: same as maxCUSize165
166
.. option:: --cutree, --no-cutree167
168
conservative, waiting until there is enough feedback in terms of 169
encoded frames to control QP. strict-cbr allows the encoder to be 170
more aggressive in hitting the target bitrate even for short segment 171
- videos. Experimental.172
+ videos.173
174
.. option:: --cbqpoffs <integer>175
176
177
undefined (not signaled)178
179
1. bt709180
- 2. undef181
+ 2. unknown182
3. **reserved**183
4. bt470m184
5. bt470bg185
186
7. smpte240m187
8. film188
9. bt2020189
+ 10. smpte428190
+ 11. smpte431191
+ 12. smpte432192
193
.. option:: --transfer <integer|string>194
195
Specify transfer characteristics. Default undefined (not signaled)196
197
1. bt709198
- 2. undef199
+ 2. unknown200
3. **reserved**201
4. bt470m202
5. bt470bg203
204
13. iec61966-2-1205
14. bt2020-10206
15. bt2020-12207
- 16. smpte-st-2084208
- 17. smpte-st-428209
+ 16. smpte2084210
+ 17. smpte428211
18. arib-std-b67212
213
.. option:: --colormatrix <integer|string>214
215
8. YCgCo216
9. bt2020nc217
10. bt2020c218
+ 11. smpte2085219
+ 12. chroma-derived-nc220
+ 13. chroma-derived-c221
+ 14. ictcp222
223
.. option:: --chromaloc <0..5>224
225
226
.. option:: --hdr, --no-hdr227
228
Force signalling of HDR parameters in SEI packets. Enabled229
- automatically when :option`--master-display` or :option`--max-cll` is230
+ automatically when :option:`--master-display` or :option:`--max-cll` is231
specified. Useful when there is a desire to signal 0 values for max-cll232
and max-fall. Default disabled.233
234
.. option:: --hdr-opt, --no-hdr-opt235
236
Add luma and chroma offsets for HDR/WCG content.237
- Input video should be 10 bit 4:2:0. Applicable for HDR content.238
- Default disabled. **Experimental Feature**239
+ Input video should be 10 bit 4:2:0. Applicable for HDR content. It is recommended240
+ that AQ-mode be enabled along with this feature. Default disabled.241
242
.. option:: --dhdr10-info <filename>243
244
245
.. option:: --min-luma <integer>246
247
Minimum luma value allowed for input pictures. Any values below min-luma248
- are clipped. Experimental. No default.249
+ are clipped. No default.250
251
.. option:: --max-luma <integer>252
253
Maximum luma value allowed for input pictures. Any values above max-luma254
- are clipped. Experimental. No default.255
+ are clipped. No default.256
257
Bitstream options258
=================259
260
.. option:: --opt-qp-pps, --no-opt-qp-pps261
262
Optimize QP in PPS (instead of default value of 26) based on the QP values263
- observed in last GOP. Default enabled.264
+ observed in last GOP. Default disabled.265
266
.. option:: --opt-ref-list-length-pps, --no-opt-ref-list-length-pps267
268
Optimize L0 and L1 ref list length in PPS (instead of default value of 0)269
- based on the lengths observed in the last GOP. Default enabled.270
+ based on the lengths observed in the last GOP. Default disabled.271
272
.. option:: --multi-pass-opt-rps, --no-multi-pass-opt-rps273
274
275
276
Only effective at RD levels 5 and 6277
278
+DCT Approximations279
+=================280
+281
+.. option:: --lowpass-dct282
+283
+ If enabled, x265 will use low-pass subband dct approximation instead of the284
+ standard dct for 16x16 and 32x32 blocks. This approximation is less computational 285
+ intensive but it generates truncated coefficient matrixes for the transformed block. 286
+ Empirical analysis shows marginal loss in compression and performance gains up to 10%,287
+ paticularly at moderate bit-rates.288
+289
+ This approximation should be considered for platforms with performance and time 290
+ constrains.291
+292
+ Default disabled. **Experimental feature**293
294
Debugging options295
=================296
x265_2.5.tar.gz/doc/reST/releasenotes.rst -> x265_2.6.tar.gz/doc/reST/releasenotes.rst
Changed
39
1
2
Release Notes3
*************4
5
+Version 2.66
+===========7
+8
+Release date - 29th November, 2017.9
+10
+New features11
+------------12
+1. x265 can now refine analysis from a previous HEVC encode (using options :option:`--refine-inter`, and :option:`--refine-intra`), or a previous AVC encode (using option :option:`--refine-mv-type`). The previous encode's information can be packaged using the *x265_analysis_data_t* data field available in the *x265_picture* object.13
+2. Basic support for segmented (or chunked) encoding added with :option:`--vbv-end` that can specify the status of CPB at the end of a segment. String this together with :option:`--vbv-init` to encode a title as chunks while maintaining VBV compliance!14
+3. :option:`--force-flush` can be used to trigger a premature flush of the encoder. This option is beneficial when input is known to be bursty, and may be at a rate slower than the encoder.15
+4. Experimental feature :option:`--lowpass-dct` that uses truncated DCT for transformation.16
+17
+Encoder enhancements18
+--------------------19
+1. Slice-parallel mode gets a significant boost in performance, particularly in low-latency mode.20
+2. x265 now officially supported on VS2017.21
+3. x265 now supports all depths from mono0 to mono16 for Y4M format.22
+23
+API changes24
+-----------25
+1. Options that modified PPS dynamically (:option:`--opt-qp-pps` and :option:`--opt-ref-list-length-pps`) are now disabled by default to enable users to save bits by not sending headers. If these options are enabled, headers have to be repeated for every GOP.26
+2. Rate-control and analysis parameters can dynamically be reconfigured simultaneously via the *x265_encoder_reconfig* API.27
+3. New API functions to extract intermediate information such as slice-type, scenecut information, reference frames, etc. are now available. This information may be beneficial to integrating applications that are attempting to perform content-adaptive encoding. Refer to documentation on *x265_get_slicetype_poc_and_scenecut*, and *x265_get_ref_frame_list* for more details and suggested usage.28
+4. A new API to pass supplemental CTU information to x265 to influence analysis decisions has been added. Refer to documentation on *x265_encoder_ctu_info* for more details.29
+30
+Bug fixes31
+---------32
+1. Bug fixes when :option:`--slices` is used with VBV settings.33
+2. Minor memory leak fixed for HDR10+ builds, and default x265 when pools option is specified.34
+3. HDR10+ bug fix to remove dependence on poc counter to select meta-data information.35
+36
Version 2.537
===========38
39
x265_2.5.tar.gz/source/CMakeLists.txt -> x265_2.6.tar.gz/source/CMakeLists.txt
Changed
58
1
2
option(STATIC_LINK_CRT "Statically link C runtime for release builds" OFF)3
mark_as_advanced(FPROFILE_USE FPROFILE_GENERATE NATIVE_BUILD)4
# X265_BUILD must be incremented each time the public API is changed5
-set(X265_BUILD 130)6
+set(X265_BUILD 146)7
configure_file("${PROJECT_SOURCE_DIR}/x265.def.in"8
"${PROJECT_BINARY_DIR}/x265.def")9
configure_file("${PROJECT_SOURCE_DIR}/x265_config.h.in"10
11
endif()12
# this option is to enable the inclusion of dynamic HDR10 library to the libx265 compilation13
option(ENABLE_HDR10_PLUS "Enable dynamic HDR10 compilation" OFF)14
+if(MSVC AND (MSVC_VERSION LESS 1800) AND ENABLE_HDR10_PLUS)15
+ message(FATAL_ERROR "MSVC version 12.0 or above required to support hdr10plus")16
+endif()17
+if(WIN32 AND (MSVC_VERSION GREATER 1800))18
+ if(CMAKE_VERSION VERSION_LESS 3.7)19
+ message(FATAL_ERROR "cmake version not compatible for VS 2017. Update the cmake to versions 3.7 or above")20
+ endif()21
+endif()22
if(GCC)23
add_definitions(-Wall -Wextra -Wshadow)24
add_definitions(-D__STDC_LIMIT_MACROS=1)25
26
endif()27
install(FILES x265.h "${PROJECT_BINARY_DIR}/x265_config.h" DESTINATION include)28
29
+if(WIN32)30
+ install(FILES "${PROJECT_BINARY_DIR}/Debug/x265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS Debug)31
+ install(FILES "${PROJECT_BINARY_DIR}/RelWithDebInfo/x265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS RelWithDebInfo)32
+ install(FILES "${PROJECT_BINARY_DIR}/Debug/libx265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS Debug OPTIONAL NAMELINK_ONLY)33
+ install(FILES "${PROJECT_BINARY_DIR}/RelWithDebInfo/libx265.pdb" DESTINATION ${BIN_INSTALL_DIR} CONFIGURATIONS RelWithDebInfo OPTIONAL NAMELINK_ONLY)34
+endif()35
+36
if(CMAKE_RC_COMPILER)37
# The resource compiler does not need CFLAGS or macro defines. It38
# often breaks them39
40
DESTINATION "${LIB_INSTALL_DIR}/pkgconfig")41
endif()42
43
-if(NOT WIN32)44
- configure_file("${CMAKE_CURRENT_SOURCE_DIR}/cmake/cmake_uninstall.cmake.in"45
- "${CMAKE_CURRENT_BINARY_DIR}/cmake/cmake_uninstall.cmake"46
- IMMEDIATE @ONLY)47
- add_custom_target(uninstall48
- "${CMAKE_COMMAND}" -P "${CMAKE_CURRENT_BINARY_DIR}/cmake/cmake_uninstall.cmake")49
-endif()50
+configure_file("${CMAKE_CURRENT_SOURCE_DIR}/cmake/cmake_uninstall.cmake.in"51
+ "${CMAKE_CURRENT_BINARY_DIR}/cmake/cmake_uninstall.cmake"52
+ IMMEDIATE @ONLY)53
+add_custom_target(uninstall54
+ "${CMAKE_COMMAND}" -P "${CMAKE_CURRENT_BINARY_DIR}/cmake/cmake_uninstall.cmake")55
56
# Main CLI application57
set(ENABLE_CLI ON CACHE BOOL "Build standalone CLI application")58
x265_2.5.tar.gz/source/cmake/cmake_uninstall.cmake.in -> x265_2.6.tar.gz/source/cmake/cmake_uninstall.cmake.in
Changed
9
1
2
message(STATUS "File '$ENV{DESTDIR}${file}' does not exist.")3
endif()4
endforeach(file)5
+6
+if(EXISTS "${CMAKE_CURRENT_BINARY_DIR}/install_manifest.txt")7
+ file(REMOVE "${CMAKE_CURRENT_BINARY_DIR}/install_manifest.txt")8
+endif()9
x265_2.5.tar.gz/source/common/CMakeLists.txt -> x265_2.6.tar.gz/source/common/CMakeLists.txt
Changed
10
1
2
add_library(common OBJECT3
${ASM_PRIMITIVES} ${VEC_PRIMITIVES} ${ALTIVEC_PRIMITIVES} ${WINXP}4
primitives.cpp primitives.h5
- pixel.cpp dct.cpp ipfilter.cpp intrapred.cpp loopfilter.cpp6
+ pixel.cpp dct.cpp lowpassdct.cpp ipfilter.cpp intrapred.cpp loopfilter.cpp7
constants.cpp constants.h8
cpu.cpp cpu.h version.cpp9
threading.cpp threading.h10
x265_2.5.tar.gz/source/common/common.h -> x265_2.6.tar.gz/source/common/common.h
Changed
9
1
2
3
// arbitrary, but low because SATD scores are 1/4 normal4
#define X265_LOOKAHEAD_QP (12 + QP_BD_OFFSET)5
-#define X265_LOOKAHEAD_MAX 2506
7
// Use the same size blocks as x264. Using larger blocks seems to give artificially8
// high cost estimates (intra and inter both suffer)9
x265_2.5.tar.gz/source/common/cudata.cpp -> x265_2.6.tar.gz/source/common/cudata.cpp
Changed
19
1
2
m_cuDepth = charBuf; charBuf += m_numPartitions;3
m_predMode = charBuf; charBuf += m_numPartitions; /* the order up to here is important in initCTU() and initSubCU() */4
m_partSize = charBuf; charBuf += m_numPartitions;5
+ m_skipFlag[0] = charBuf; charBuf += m_numPartitions;6
+ m_skipFlag[1] = charBuf; charBuf += m_numPartitions;7
m_mergeFlag = charBuf; charBuf += m_numPartitions;8
m_interDir = charBuf; charBuf += m_numPartitions;9
m_mvpIdx[0] = charBuf; charBuf += m_numPartitions;10
11
m_cuDepth = charBuf; charBuf += m_numPartitions;12
m_predMode = charBuf; charBuf += m_numPartitions; /* the order up to here is important in initCTU() and initSubCU() */13
m_partSize = charBuf; charBuf += m_numPartitions;14
+ m_skipFlag[0] = charBuf; charBuf += m_numPartitions;15
+ m_skipFlag[1] = charBuf; charBuf += m_numPartitions;16
m_mergeFlag = charBuf; charBuf += m_numPartitions;17
m_interDir = charBuf; charBuf += m_numPartitions;18
m_mvpIdx[0] = charBuf; charBuf += m_numPartitions;19
x265_2.5.tar.gz/source/common/cudata.h -> x265_2.6.tar.gz/source/common/cudata.h
Changed
17
1
2
uint8_t* m_predMode; // array of prediction modes3
uint8_t* m_partSize; // array of partition sizes4
uint8_t* m_mergeFlag; // array of merge flags5
+ uint8_t* m_skipFlag[2];6
uint8_t* m_interDir; // array of inter directions7
uint8_t* m_mvpIdx[2]; // array of motion vector predictor candidates or merge candidate indices [0]8
uint8_t* m_tuDepth; // array of transform indices9
uint8_t* m_transformSkip[3]; // array of transform skipping flags per plane10
uint8_t* m_cbf[3]; // array of coded block flags (CBF) per plane11
uint8_t* m_chromaIntraDir; // array of intra directions (chroma)12
- enum { BytesPerPartition = 21 }; // combined sizeof() of all per-part data13
+ enum { BytesPerPartition = 23 }; // combined sizeof() of all per-part data14
15
sse_t* m_distortion;16
coeff_t* m_trCoeff[3]; // transformed coefficient buffer per plane17
x265_2.5.tar.gz/source/common/frame.cpp -> x265_2.6.tar.gz/source/common/frame.cpp
Changed
28
1
2
}3
}4
5
- if (m_fencPic->create(param) && m_lowres.create(m_fencPic, param->bframes, !!param->rc.aqMode || !!param->bAQMotion, param->rc.qgSize))6
+ if (param->bMVType == AVC_INFO)7
+ {8
+ m_analysisData.wt = NULL;9
+ m_analysisData.intraData = NULL;10
+ m_analysisData.interData = NULL;11
+ m_analysis2Pass.analysisFramedata = NULL;12
+ }13
+14
+ if (m_fencPic->create(param, !!m_param->bCopyPicToFrame) && m_lowres.create(m_fencPic, param->bframes, !!param->rc.aqMode || !!param->bAQMotion, param->rc.qgSize))15
{16
X265_CHECK((m_reconColCount == NULL), "m_reconColCount was initialized");17
m_numRows = (m_fencPic->m_picHeight + param->maxCUSize - 1) / param->maxCUSize;18
19
20
if (m_fencPic)21
{22
- m_fencPic->destroy();23
+ if (m_param->bCopyPicToFrame)24
+ m_fencPic->destroy();25
delete m_fencPic;26
m_fencPic = NULL;27
}28
x265_2.5.tar.gz/source/common/frame.h -> x265_2.6.tar.gz/source/common/frame.h
Changed
18
1
2
3
float* m_quantOffsets; // points to quantOffsets in x265_picture4
x265_sei m_userSEI;5
+ Event m_reconEncoded;6
7
/* Frame Parallelism - notification between FrameEncoders of available motion reference rows */8
ThreadSafeInteger* m_reconRowFlag; // flag of CTU rows completely reconstructed and extended for motion reference9
10
x265_analysis_2Pass m_analysis2Pass;11
RcStats* m_rcData;12
13
+ Event m_copyMVType;14
+15
x265_ctu_info_t** m_ctuInfo;16
Event m_copied;17
int* m_prevCtuInfoChange;18
x265_2.5.tar.gz/source/common/framedata.h -> x265_2.6.tar.gz/source/common/framedata.h
Changed
9
1
2
uint8_t* mvpIdx[2];3
int8_t* refIdx[2];4
MV* mv[2];5
+ int64_t* sadCost;6
};7
8
struct analysis2PassFrameData9
x265_2.6.tar.gz/source/common/lowpassdct.cpp
Added
129
1
2
+/*****************************************************************************3
+ * Copyright (C) 2017 4
+ *5
+ * Authors: Humberto Ribeiro Filho <mont3z.claro5@gmail.com>6
+ *7
+ * This program is free software; you can redistribute it and/or modify8
+ * it under the terms of the GNU General Public License as published by9
+ * the Free Software Foundation; either version 2 of the License, or10
+ * (at your option) any later version.11
+ *12
+ * This program is distributed in the hope that it will be useful,13
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of14
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the15
+ * GNU General Public License for more details.16
+ *17
+ * You should have received a copy of the GNU General Public License18
+ * along with this program; if not, write to the Free Software19
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.20
+ *21
+ * This program is also available under a commercial proprietary license.22
+ * For more information, contact us at license @ x265.com.23
+ *****************************************************************************/24
+25
+#include "common.h"26
+#include "primitives.h"27
+28
+using namespace X265_NS;29
+30
+/* standard dct transformations */31
+static dct_t* s_dct4x4;32
+static dct_t* s_dct8x8;33
+static dct_t* s_dct16x16;34
+35
+static void lowPassDct8_c(const int16_t* src, int16_t* dst, intptr_t srcStride)36
+{37
+ ALIGN_VAR_32(int16_t, coef[4 * 4]);38
+ ALIGN_VAR_32(int16_t, avgBlock[4 * 4]);39
+ int16_t totalSum = 0;40
+ int16_t sum = 0;41
+ 42
+ for (int i = 0; i < 4; i++)43
+ for (int j =0; j < 4; j++)44
+ {45
+ // Calculate average of 2x2 cells46
+ sum = src[2*i*srcStride + 2*j] + src[2*i*srcStride + 2*j + 1]47
+ + src[(2*i+1)*srcStride + 2*j] + src[(2*i+1)*srcStride + 2*j + 1];48
+ avgBlock[i*4 + j] = sum >> 2;49
+50
+ totalSum += sum; // use to calculate total block average51
+ }52
+53
+ //dct454
+ (*s_dct4x4)(avgBlock, coef, 4);55
+ memset(dst, 0, 64 * sizeof(int16_t));56
+ for (int i = 0; i < 4; i++)57
+ {58
+ memcpy(&dst[i * 8], &coef[i * 4], 4 * sizeof(int16_t));59
+ }60
+61
+ // replace first coef with total block average62
+ dst[0] = totalSum << 1;63
+}64
+65
+static void lowPassDct16_c(const int16_t* src, int16_t* dst, intptr_t srcStride)66
+{67
+ ALIGN_VAR_32(int16_t, coef[8 * 8]);68
+ ALIGN_VAR_32(int16_t, avgBlock[8 * 8]);69
+ int32_t totalSum = 0;70
+ int16_t sum = 0;71
+ for (int i = 0; i < 8; i++)72
+ for (int j =0; j < 8; j++)73
+ {74
+ sum = src[2*i*srcStride + 2*j] + src[2*i*srcStride + 2*j + 1]75
+ + src[(2*i+1)*srcStride + 2*j] + src[(2*i+1)*srcStride + 2*j + 1];76
+ avgBlock[i*8 + j] = sum >> 2;77
+78
+ totalSum += sum;79
+ }80
+81
+ (*s_dct8x8)(avgBlock, coef, 8);82
+ memset(dst, 0, 256 * sizeof(int16_t));83
+ for (int i = 0; i < 8; i++)84
+ {85
+ memcpy(&dst[i * 16], &coef[i * 8], 8 * sizeof(int16_t));86
+ }87
+ dst[0] = static_cast<int16_t>(totalSum >> 1);88
+}89
+90
+static void lowPassDct32_c(const int16_t* src, int16_t* dst, intptr_t srcStride)91
+{92
+ ALIGN_VAR_32(int16_t, coef[16 * 16]);93
+ ALIGN_VAR_32(int16_t, avgBlock[16 * 16]);94
+ int32_t totalSum = 0;95
+ int16_t sum = 0;96
+ for (int i = 0; i < 16; i++)97
+ for (int j =0; j < 16; j++)98
+ {99
+ sum = src[2*i*srcStride + 2*j] + src[2*i*srcStride + 2*j + 1]100
+ + src[(2*i+1)*srcStride + 2*j] + src[(2*i+1)*srcStride + 2*j + 1];101
+ avgBlock[i*16 + j] = sum >> 2;102
+103
+ totalSum += sum;104
+ }105
+106
+ (*s_dct16x16)(avgBlock, coef, 16);107
+ memset(dst, 0, 1024 * sizeof(int16_t));108
+ for (int i = 0; i < 16; i++)109
+ {110
+ memcpy(&dst[i * 32], &coef[i * 16], 16 * sizeof(int16_t));111
+ }112
+ dst[0] = static_cast<int16_t>(totalSum >> 3);113
+}114
+115
+namespace X265_NS {116
+// x265 private namespace117
+118
+void setupLowPassPrimitives_c(EncoderPrimitives& p)119
+{120
+ s_dct4x4 = &(p.cu[BLOCK_4x4].standard_dct);121
+ s_dct8x8 = &(p.cu[BLOCK_8x8].standard_dct);122
+ s_dct16x16 = &(p.cu[BLOCK_16x16].standard_dct);123
+124
+ p.cu[BLOCK_8x8].lowpass_dct = lowPassDct8_c;125
+ p.cu[BLOCK_16x16].lowpass_dct = lowPassDct16_c;126
+ p.cu[BLOCK_32x32].lowpass_dct = lowPassDct32_c;127
+}128
+}129
x265_2.5.tar.gz/source/common/lowres.cpp -> x265_2.6.tar.gz/source/common/lowres.cpp
Changed
11
1
2
3
for (int i = 0; i < bframes + 2; i++)4
intraMbs[i] = 0;5
+ if (origPic->m_param->rc.vbvBufferSize)6
+ for (int i = 0; i < X265_LOOKAHEAD_MAX + 1; i++)7
+ plannedType[i] = X265_TYPE_AUTO;8
9
/* downscale and generate 4 hpel planes for lookahead */10
primitives.frameInitLowres(origPic->m_picOrg[0],11
x265_2.5.tar.gz/source/common/param.cpp -> x265_2.6.tar.gz/source/common/param.cpp
Changed
190
1
2
param->bEnableConstrainedIntra = 0;3
param->bEnableStrongIntraSmoothing = 1;4
param->bEnableFastIntra = 0;5
+ param->bEnableSplitRdSkip = 0;6
7
/* Inter Coding tools */8
param->searchMethod = X265_HEX_SEARCH;9
10
param->rc.vbvMaxBitrate = 0;11
param->rc.vbvBufferSize = 0;12
param->rc.vbvBufferInit = 0.9;13
+ param->vbvBufferEnd = 0;14
+ param->vbvEndFrameAdjust = 0;15
param->rc.rfConstant = 28;16
param->rc.bitrate = 0;17
param->rc.qCompress = 0.6;18
19
20
param->bEmitVUITimingInfo = 1;21
param->bEmitVUIHRDInfo = 1;22
- param->bOptQpPPS = 1;23
- param->bOptRefListLengthPPS = 1;24
+ param->bOptQpPPS = 0;25
+ param->bOptRefListLengthPPS = 0;26
param->bOptCUDeltaQP = 0;27
param->bAQMotion = 0;28
param->bHDROpt = 0;29
30
param->mvRefine = 0;31
param->bUseAnalysisFile = 1;32
param->csvfpt = NULL;33
+ param->forceFlush = 0;34
+ param->bDisableLookahead = 0;35
+ param->bCopyPicToFrame = 1;36
+37
+ /* DCT Approximations */38
+ param->bLowPassDct = 0;39
+ param->bMVType = 0;40
}41
42
int x265_param_default_preset(x265_param* param, const char* preset, const char* tune)43
44
OPT("ctu-info") p->bCTUInfo = atoi(value);45
OPT("scale-factor") p->scaleFactor = atoi(value);46
OPT("refine-intra")p->intraRefine = atoi(value);47
- OPT("refine-inter")p->interRefine = atobool(value);48
+ OPT("refine-inter")p->interRefine = atoi(value);49
OPT("refine-mv")p->mvRefine = atobool(value);50
+ OPT("force-flush")p->forceFlush = atoi(value);51
+ OPT("splitrd-skip") p->bEnableSplitRdSkip = atobool(value);52
+ OPT("lowpass-dct") p->bLowPassDct = atobool(value);53
+ OPT("vbv-end") p->vbvBufferEnd = atof(value);54
+ OPT("vbv-end-fr-adj") p->vbvEndFrameAdjust = atof(value);55
+ OPT("copy-pic") p->bCopyPicToFrame = atobool(value);56
+ OPT("refine-mv-type")57
+ {58
+ if (strcmp(strdup(value), "avc") == 0)59
+ {60
+ p->bMVType = AVC_INFO;61
+ }62
+ else if (strcmp(strdup(value), "off") == 0)63
+ {64
+ p->bMVType = NO_INFO;65
+ }66
+ else67
+ {68
+ bError = true;69
+ }70
+ }71
else72
return X265_PARAM_BAD_NAME;73
}74
75
"Video Format must be component,"76
" pal, ntsc, secam, mac or undef");77
CHECK(param->vui.colorPrimaries < 078
- || param->vui.colorPrimaries > 979
+ || param->vui.colorPrimaries > 1280
|| param->vui.colorPrimaries == 3,81
"Color Primaries must be undef, bt709, bt470m,"82
- " bt470bg, smpte170m, smpte240m, film or bt2020");83
+ " bt470bg, smpte170m, smpte240m, film, bt2020, smpte-st-428, smpte-rp-431 or smpte-eg-432");84
CHECK(param->vui.transferCharacteristics < 085
|| param->vui.transferCharacteristics > 1886
|| param->vui.transferCharacteristics == 3,87
88
" smpte170m, smpte240m, linear, log100, log316, iec61966-2-4, bt1361e,"89
" iec61966-2-1, bt2020-10, bt2020-12, smpte-st-2084, smpte-st-428 or arib-std-b67");90
CHECK(param->vui.matrixCoeffs < 091
- || param->vui.matrixCoeffs > 1092
+ || param->vui.matrixCoeffs > 1493
|| param->vui.matrixCoeffs == 3,94
"Matrix Coefficients must be undef, bt709, fcc, bt470bg, smpte170m,"95
- " smpte240m, GBR, YCgCo, bt2020nc or bt2020c");96
+ " smpte240m, GBR, YCgCo, bt2020nc, bt2020c, smpte-st-2085, chroma-nc, chroma-c or ictcp");97
CHECK(param->vui.chromaSampleLocTypeTopField < 098
|| param->vui.chromaSampleLocTypeTopField > 5,99
"Chroma Sample Location Type Top Field must be 0-5");100
101
"Maximum local bit rate can not be less than zero");102
CHECK(param->rc.vbvBufferInit < 0,103
"Valid initial VBV buffer occupancy must be a fraction 0 - 1, or size in kbits");104
+ CHECK(param->vbvBufferEnd < 0,105
+ "Valid final VBV buffer emptiness must be a fraction 0 - 1, or size in kbits");106
+ CHECK(param->vbvEndFrameAdjust < 0,107
+ "Valid vbv-end-fr-adj must be a fraction 0 - 1");108
+ CHECK(!param->totalFrames && param->vbvEndFrameAdjust,109
+ "vbv-end-fr-adj cannot be enabled when total number of frames is unknown");110
CHECK(param->rc.bitrate < 0,111
"Target bitrate can not be less than zero");112
CHECK(param->rc.qCompress < 0.5 || param->rc.qCompress > 1.0,113
114
"Supported range for log2MaxPocLsb is 4 to 16");115
CHECK(param->bCTUInfo < 0 || (param->bCTUInfo != 0 && param->bCTUInfo != 1 && param->bCTUInfo != 2 && param->bCTUInfo != 4 && param->bCTUInfo != 6) || param->bCTUInfo > 6,116
"Supported values for bCTUInfo are 0, 1, 2, 4, 6");117
+ CHECK(param->interRefine > 3 || param->interRefine < 0,118
+ "Invalid refine-inter value, refine-inter levels 0 to 3 supported");119
+ CHECK(param->intraRefine > 3 || param->intraRefine < 0,120
+ "Invalid refine-intra value, refine-intra levels 0 to 3 supported");121
#if !X86_64122
CHECK(param->searchMethod == X265_SEA && (param->sourceWidth > 840 || param->sourceHeight > 480),123
"SEA motion search does not support resolutions greater than 480p in 32 bit build");124
125
}126
127
if (param->rc.vbvBufferSize)128
- x265_log(param, X265_LOG_INFO, "VBV/HRD buffer / max-rate / init : %d / %d / %.3f\n",129
- param->rc.vbvBufferSize, param->rc.vbvMaxBitrate, param->rc.vbvBufferInit);130
-131
+ {132
+ if (param->vbvBufferEnd)133
+ x265_log(param, X265_LOG_INFO, "VBV/HRD buffer / max-rate / init / end / fr-adj: %d / %d / %.3f / %.3f / %.3f\n",134
+ param->rc.vbvBufferSize, param->rc.vbvMaxBitrate, param->rc.vbvBufferInit, param->vbvBufferEnd, param->vbvEndFrameAdjust);135
+ else136
+ x265_log(param, X265_LOG_INFO, "VBV/HRD buffer / max-rate / init : %d / %d / %.3f\n",137
+ param->rc.vbvBufferSize, param->rc.vbvMaxBitrate, param->rc.vbvBufferInit);138
+ }139
+ 140
char buf[80] = { 0 };141
char tmp[40];142
#define TOOLOPT(FLAG, STR) if (FLAG) appendtool(param, buf, sizeof(buf), STR);143
144
TOOLOPT(param->bEnableRdRefine, "rd-refine");145
TOOLOPT(param->bEnableEarlySkip, "early-skip");146
TOOLOPT(param->bEnableRecursionSkip, "rskip");147
+ TOOLOPT(param->bEnableSplitRdSkip, "splitrd-skip");148
TOOLVAL(param->noiseReductionIntra, "nr-intra=%d");149
TOOLVAL(param->noiseReductionInter, "nr-inter=%d");150
TOOLOPT(param->bEnableTSkipFast, "tskip-fast");151
152
TOOLVAL(param->lookaheadSlices, "lslices=%d");153
TOOLVAL(param->lookaheadThreads, "lthreads=%d")154
TOOLVAL(param->bCTUInfo, "ctu-info=%d");155
+ if (param->bMVType == AVC_INFO)156
+ TOOLOPT(param->bMVType, "refine-mv-type=avc");157
if (param->maxSlices > 1)158
TOOLVAL(param->maxSlices, "slices=%d");159
if (param->bEnableLoopFilter)160
161
BOOL(p->bEnableTSkipFast, "tskip-fast");162
BOOL(p->bCULossless, "cu-lossless");163
BOOL(p->bIntraInBFrames, "b-intra");164
+ BOOL(p->bEnableSplitRdSkip, "splitrd-skip");165
s += sprintf(s, " rdpenalty=%d", p->rdPenalty);166
s += sprintf(s, " psy-rd=%.2f", p->psyRd);167
s += sprintf(s, " psy-rdoq=%.2f", p->psyRdoq);168
169
{170
s += sprintf(s, " vbv-maxrate=%d vbv-bufsize=%d vbv-init=%.1f",171
p->rc.vbvMaxBitrate, p->rc.vbvBufferSize, p->rc.vbvBufferInit);172
+ if (p->vbvBufferEnd)173
+ s += sprintf(s, " vbv-end=%.1f vbv-end-fr-adj=%.1f", p->vbvBufferEnd, p->vbvEndFrameAdjust);174
if (p->rc.rateControlMode == X265_RC_CRF)175
- s += sprintf(s, " crf-max=%.1f crf-min=%.1f", p->rc.rfConstantMax, p->rc.rfConstantMin);176
+ s += sprintf(s, " crf-max=%.1f crf-min=%.1f", p->rc.rfConstantMax, p->rc.rfConstantMin); 177
}178
}179
else if (p->rc.rateControlMode == X265_RC_CQP)180
181
s += sprintf(s, " refine-mv=%d", p->mvRefine);182
BOOL(p->bLimitSAO, "limit-sao");183
s += sprintf(s, " ctu-info=%d", p->bCTUInfo);184
+ BOOL(p->bLowPassDct, "lowpass-dct");185
+ s += sprintf(s, " refine-mv-type=%d", p->bMVType);186
+ s += sprintf(s, " copy-pic=%d", p->bCopyPicToFrame);187
#undef BOOL188
return buf;189
}190
x265_2.5.tar.gz/source/common/piclist.cpp -> x265_2.6.tar.gz/source/common/piclist.cpp
Changed
17
1
2
return NULL;3
}4
5
+Frame* PicList::getCurFrame(void)6
+{7
+ Frame *curFrame = m_start;8
+ if (curFrame != NULL)9
+ return curFrame;10
+ else11
+ return NULL;12
+}13
+14
void PicList::remove(Frame& curFrame)15
{16
#if _DEBUG17
x265_2.5.tar.gz/source/common/piclist.h -> x265_2.6.tar.gz/source/common/piclist.h
Changed
11
1
2
/** Find frame with specified POC */3
Frame* getPOC(int poc);4
5
+ /** Get the current Frame from the list **/6
+ Frame* getCurFrame(void);7
+8
/** Remove picture from list */9
void remove(Frame& pic);10
11
x265_2.5.tar.gz/source/common/picyuv.cpp -> x265_2.6.tar.gz/source/common/picyuv.cpp
Changed
149
1
2
m_vChromaShift = 0;3
}4
5
-bool PicYuv::create(x265_param* param, pixel *pixelbuf)6
+bool PicYuv::create(x265_param* param, bool picAlloc, pixel *pixelbuf)7
{8
m_param = param;9
uint32_t picWidth = m_param->sourceWidth;10
11
m_picOrg[0] = pixelbuf;12
else13
{14
- CHECKED_MALLOC(m_picBuf[0], pixel, m_stride * (maxHeight + (m_lumaMarginY * 2)));15
- m_picOrg[0] = m_picBuf[0] + m_lumaMarginY * m_stride + m_lumaMarginX;16
+ if (picAlloc)17
+ {18
+ CHECKED_MALLOC(m_picBuf[0], pixel, m_stride * (maxHeight + (m_lumaMarginY * 2)));19
+ m_picOrg[0] = m_picBuf[0] + m_lumaMarginY * m_stride + m_lumaMarginX;20
+ }21
}22
23
if (picCsp != X265_CSP_I400)24
25
m_chromaMarginX = m_lumaMarginX; // keep 16-byte alignment for chroma CTUs26
m_chromaMarginY = m_lumaMarginY >> m_vChromaShift;27
m_strideC = ((numCuInWidth * m_param->maxCUSize) >> m_hChromaShift) + (m_chromaMarginX * 2);28
+ if (picAlloc)29
+ {30
+ CHECKED_MALLOC(m_picBuf[1], pixel, m_strideC * ((maxHeight >> m_vChromaShift) + (m_chromaMarginY * 2)));31
+ CHECKED_MALLOC(m_picBuf[2], pixel, m_strideC * ((maxHeight >> m_vChromaShift) + (m_chromaMarginY * 2)));32
33
- CHECKED_MALLOC(m_picBuf[1], pixel, m_strideC * ((maxHeight >> m_vChromaShift) + (m_chromaMarginY * 2)));34
- CHECKED_MALLOC(m_picBuf[2], pixel, m_strideC * ((maxHeight >> m_vChromaShift) + (m_chromaMarginY * 2)));35
-36
- m_picOrg[1] = m_picBuf[1] + m_chromaMarginY * m_strideC + m_chromaMarginX;37
- m_picOrg[2] = m_picBuf[2] + m_chromaMarginY * m_strideC + m_chromaMarginX;38
+ m_picOrg[1] = m_picBuf[1] + m_chromaMarginY * m_strideC + m_chromaMarginX;39
+ m_picOrg[2] = m_picBuf[2] + m_chromaMarginY * m_strideC + m_chromaMarginX;40
+ }41
}42
else43
{44
45
uint64_t crSum;46
lumaSum = cbSum = crSum = 0;47
48
- if (pic.bitDepth == 8)49
+ if (m_param->bCopyPicToFrame)50
{51
+ if (pic.bitDepth == 8)52
+ {53
#if (X265_DEPTH > 8)54
{55
pixel *yPixel = m_picOrg[0];56
57
}58
}59
#else /* Case for (X265_DEPTH == 8) */60
- // TODO: Does we need this path? may merge into above in future61
+ // TODO: Does we need this path? may merge into above in future62
{63
pixel *yPixel = m_picOrg[0];64
uint8_t *yChar = (uint8_t*)pic.planes[0];65
66
}67
}68
#endif /* (X265_DEPTH > 8) */69
- }70
- else /* pic.bitDepth > 8 */71
- {72
- /* defensive programming, mask off bits that are supposed to be zero */73
- uint16_t mask = (1 << X265_DEPTH) - 1;74
- int shift = abs(pic.bitDepth - X265_DEPTH);75
- pixel *yPixel = m_picOrg[0];76
-77
- uint16_t *yShort = (uint16_t*)pic.planes[0];78
-79
- if (pic.bitDepth > X265_DEPTH)80
- {81
- /* shift right and mask pixels to final size */82
- primitives.planecopy_sp(yShort, pic.stride[0] / sizeof(*yShort), yPixel, m_stride, width, height, shift, mask);83
- }84
- else /* Case for (pic.bitDepth <= X265_DEPTH) */85
- {86
- /* shift left and mask pixels to final size */87
- primitives.planecopy_sp_shl(yShort, pic.stride[0] / sizeof(*yShort), yPixel, m_stride, width, height, shift, mask);88
}89
-90
- if (param.internalCsp != X265_CSP_I400)91
+ else /* pic.bitDepth > 8 */92
{93
- pixel *uPixel = m_picOrg[1];94
- pixel *vPixel = m_picOrg[2];95
+ /* defensive programming, mask off bits that are supposed to be zero */96
+ uint16_t mask = (1 << X265_DEPTH) - 1;97
+ int shift = abs(pic.bitDepth - X265_DEPTH);98
+ pixel *yPixel = m_picOrg[0];99
100
- uint16_t *uShort = (uint16_t*)pic.planes[1];101
- uint16_t *vShort = (uint16_t*)pic.planes[2];102
+ uint16_t *yShort = (uint16_t*)pic.planes[0];103
104
if (pic.bitDepth > X265_DEPTH)105
{106
- primitives.planecopy_sp(uShort, pic.stride[1] / sizeof(*uShort), uPixel, m_strideC, width >> m_hChromaShift, height >> m_vChromaShift, shift, mask);107
- primitives.planecopy_sp(vShort, pic.stride[2] / sizeof(*vShort), vPixel, m_strideC, width >> m_hChromaShift, height >> m_vChromaShift, shift, mask);108
+ /* shift right and mask pixels to final size */109
+ primitives.planecopy_sp(yShort, pic.stride[0] / sizeof(*yShort), yPixel, m_stride, width, height, shift, mask);110
}111
else /* Case for (pic.bitDepth <= X265_DEPTH) */112
{113
- primitives.planecopy_sp_shl(uShort, pic.stride[1] / sizeof(*uShort), uPixel, m_strideC, width >> m_hChromaShift, height >> m_vChromaShift, shift, mask);114
- primitives.planecopy_sp_shl(vShort, pic.stride[2] / sizeof(*vShort), vPixel, m_strideC, width >> m_hChromaShift, height >> m_vChromaShift, shift, mask);115
+ /* shift left and mask pixels to final size */116
+ primitives.planecopy_sp_shl(yShort, pic.stride[0] / sizeof(*yShort), yPixel, m_stride, width, height, shift, mask);117
+ }118
+119
+ if (param.internalCsp != X265_CSP_I400)120
+ {121
+ pixel *uPixel = m_picOrg[1];122
+ pixel *vPixel = m_picOrg[2];123
+124
+ uint16_t *uShort = (uint16_t*)pic.planes[1];125
+ uint16_t *vShort = (uint16_t*)pic.planes[2];126
+127
+ if (pic.bitDepth > X265_DEPTH)128
+ {129
+ primitives.planecopy_sp(uShort, pic.stride[1] / sizeof(*uShort), uPixel, m_strideC, width >> m_hChromaShift, height >> m_vChromaShift, shift, mask);130
+ primitives.planecopy_sp(vShort, pic.stride[2] / sizeof(*vShort), vPixel, m_strideC, width >> m_hChromaShift, height >> m_vChromaShift, shift, mask);131
+ }132
+ else /* Case for (pic.bitDepth <= X265_DEPTH) */133
+ {134
+ primitives.planecopy_sp_shl(uShort, pic.stride[1] / sizeof(*uShort), uPixel, m_strideC, width >> m_hChromaShift, height >> m_vChromaShift, shift, mask);135
+ primitives.planecopy_sp_shl(vShort, pic.stride[2] / sizeof(*vShort), vPixel, m_strideC, width >> m_hChromaShift, height >> m_vChromaShift, shift, mask);136
+ }137
}138
}139
}140
+ else141
+ {142
+ m_picOrg[0] = (pixel*)pic.planes[0];143
+ m_picOrg[1] = (pixel*)pic.planes[1];144
+ m_picOrg[2] = (pixel*)pic.planes[2];145
+ }146
147
pixel *Y = m_picOrg[0];148
pixel *U = m_picOrg[1];149
x265_2.5.tar.gz/source/common/picyuv.h -> x265_2.6.tar.gz/source/common/picyuv.h
Changed
27
1
2
#include "common.h"3
#include "md5.h"4
#include "x265.h"5
+struct x265_picyuv {};6
7
namespace X265_NS {8
// private namespace9
10
class ShortYuv;11
struct SPS;12
13
-class PicYuv14
+class PicYuv : public x265_picyuv15
{16
public:17
18
19
20
PicYuv();21
22
- bool create(x265_param* param, pixel *pixelbuf = NULL);23
+ bool create(x265_param* param, bool picAlloc = true, pixel *pixelbuf = NULL);24
bool createOffsets(const SPS& sps);25
void destroy();26
int getLumaBufLen(uint32_t picWidth, uint32_t picHeight, uint32_t picCsp);27
x265_2.5.tar.gz/source/common/primitives.cpp -> x265_2.6.tar.gz/source/common/primitives.cpp
Changed
47
1
2
void setupLoopFilterPrimitives_c(EncoderPrimitives &p);3
void setupSaoPrimitives_c(EncoderPrimitives &p);4
void setupSeaIntegralPrimitives_c(EncoderPrimitives &p);5
+void setupLowPassPrimitives_c(EncoderPrimitives& p);6
7
void setupCPrimitives(EncoderPrimitives &p)8
{9
setupPixelPrimitives_c(p); // pixel.cpp10
setupDCTPrimitives_c(p); // dct.cpp11
+ setupLowPassPrimitives_c(p); // lowpassdct.cpp12
setupFilterPrimitives_c(p); // ipfilter.cpp13
setupIntraPrimitives_c(p); // intrapred.cpp14
setupLoopFilterPrimitives_c(p); // loopfilter.cpp15
16
setupSeaIntegralPrimitives_c(p); // framefilter.cpp17
}18
19
+void enableLowpassDCTPrimitives(EncoderPrimitives &p)20
+{21
+ // update copies of the standard dct transform22
+ p.cu[BLOCK_4x4].standard_dct = p.cu[BLOCK_4x4].dct;23
+ p.cu[BLOCK_8x8].standard_dct = p.cu[BLOCK_8x8].dct;24
+ p.cu[BLOCK_16x16].standard_dct = p.cu[BLOCK_16x16].dct;25
+ p.cu[BLOCK_32x32].standard_dct = p.cu[BLOCK_32x32].dct;26
+27
+ // replace active dct by lowpass dct for high dct transforms28
+ p.cu[BLOCK_16x16].dct = p.cu[BLOCK_16x16].lowpass_dct;29
+ p.cu[BLOCK_32x32].dct = p.cu[BLOCK_32x32].lowpass_dct;30
+}31
+32
void setupAliasPrimitives(EncoderPrimitives &p)33
{34
#if HIGH_BIT_DEPTH35
36
#endif37
38
setupAliasPrimitives(primitives);39
+40
+ if (param->bLowPassDct)41
+ {42
+ enableLowpassDCTPrimitives(primitives); 43
+ }44
}45
46
x265_report_simd(param);47
x265_2.5.tar.gz/source/common/primitives.h -> x265_2.6.tar.gz/source/common/primitives.h
Changed
16
1
2
* primitives will leave 64x64 pointers NULL. Indexed by LumaCU */3
struct CU4
{5
- dct_t dct;6
- idct_t idct;7
+ dct_t dct; // active dct transformation8
+ idct_t idct; // active idct transformation9
+10
+ dct_t standard_dct; // original dct function, used by lowpass_dct11
+ dct_t lowpass_dct; // lowpass dct approximation12
+13
calcresidual_t calcresidual;14
pixel_sub_ps_t sub_ps;15
pixel_add_ps_t add_ps;16
x265_2.5.tar.gz/source/common/threadpool.cpp -> x265_2.6.tar.gz/source/common/threadpool.cpp
Changed
9
1
2
if ((nodeMaskPerPool[node] >> j) & 1)3
len += sprintf(nodesstr + len, ",%d", j);4
x265_log(p, X265_LOG_INFO, "Thread pool %d using %d threads on numa nodes %s\n", i, numThreads, nodesstr + 1);5
+ delete[] nodesstr;6
}7
else8
x265_log(p, X265_LOG_INFO, "Thread pool created using %d threads\n", numThreads);9
x265_2.5.tar.gz/source/common/wavefront.cpp -> x265_2.6.tar.gz/source/common/wavefront.cpp
Changed
19
1
2
if (m_externalDependencyBitmap)3
memset((void*)m_externalDependencyBitmap, 0, sizeof(uint32_t) * m_numWords);4
5
+ m_row_to_idx = X265_MALLOC(uint32_t, m_numRows);6
+ m_idx_to_row = X265_MALLOC(uint32_t, m_numRows);7
+8
return m_internalDependencyBitmap && m_externalDependencyBitmap;9
}10
11
WaveFront::~WaveFront()12
{13
+ x265_free((void*)m_row_to_idx);14
+ x265_free((void*)m_idx_to_row);15
+16
x265_free((void*)m_internalDependencyBitmap);17
x265_free((void*)m_externalDependencyBitmap);18
}19
x265_2.5.tar.gz/source/common/wavefront.h -> x265_2.6.tar.gz/source/common/wavefront.h
Changed
12
1
2
3
int m_numRows;4
5
+protected:6
+ uint32_t *m_row_to_idx;7
+ uint32_t *m_idx_to_row;8
+9
public:10
11
WaveFront()12
x265_2.5.tar.gz/source/dynamicHDR10/BasicStructures.h -> x265_2.6.tar.gz/source/dynamicHDR10/BasicStructures.h
Changed
38
1
2
/**3
- * @file BasicStructures.h4
- * @brief Defines the structure of metadata parameters5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,30
- * MA 02110-1301, USA.31
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.32
+ *33
+ * This program is also available under a commercial proprietary license.34
+ * For more information, contact us at license @ x265.com.35
**/36
37
#ifndef BASICSTRUCTURES_H38
x265_2.5.tar.gz/source/dynamicHDR10/JsonHelper.cpp -> x265_2.6.tar.gz/source/dynamicHDR10/JsonHelper.cpp
Changed
57
1
2
/**3
- * @file JsonHelper.cpp4
- * @brief Helper class for JSON parsing5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,30
- * MA 02110-1301, USA.31
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.32
+ *33
+ * This program is also available under a commercial proprietary license.34
+ * For more information, contact us at license @ x265.com.35
**/36
37
#include "JsonHelper.h"38
39
40
tfile.close();41
42
- size_t beginning = json_str2.find_first_of("[");43
- int fixchar = json_str2[json_str2.size() - 2] == ']' ? 1 : 0;44
- return Json::parse(json_str2.substr(beginning,json_str2.size() - fixchar),err).array_items();45
+ vector<Json> data;46
+ if (json_str2.size() != 0)47
+ {48
+ size_t beginning = json_str2.find_first_of("[");49
+ int fixchar = json_str2[json_str2.size() - 2] == ']' ? 1 : 0;50
+ return Json::parse(json_str2.substr(beginning, json_str2.size() - fixchar), err).array_items();51
+ }52
+ else53
+ return data;54
}55
56
bool JsonHelper::validatePathExtension(string &path)57
x265_2.5.tar.gz/source/dynamicHDR10/JsonHelper.h -> x265_2.6.tar.gz/source/dynamicHDR10/JsonHelper.h
Changed
38
1
2
/**3
- * @file JsonHelper.h4
- * @brief Helper class for JSON parsing5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,30
- * MA 02110-1301, USA.31
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.32
+ *33
+ * This program is also available under a commercial proprietary license.34
+ * For more information, contact us at license @ x265.com.35
**/36
37
#ifndef JSON_H38
x265_2.5.tar.gz/source/dynamicHDR10/LICENSE.txt -> x265_2.6.tar.gz/source/dynamicHDR10/LICENSE.txt
Changed
31
1
2
-Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana3
+Copyright (C) 2013-2017 MulticoreWare, Inc4
5
-This software is the confidential and proprietary information of Samsung Electronics, Inc. ("Confidential Information").6
-You shall not disclose such Confidential Information and shall use it only in accordance with the terms of the license agreement7
-you entered into with Samsung.8
-9
-This program is free software; you can redistribute it and/or10
-modify it under the terms of the GNU General Public License11
-as published by the Free Software Foundation; either version 212
-of the License, or (at your option) any later version.13
+This program is free software; you can redistribute it and/or modify14
+it under the terms of the GNU General Public License as published by15
+the Free Software Foundation; either version 2 of the License, or16
+(at your option) any later version.17
18
This program is distributed in the hope that it will be useful,19
but WITHOUT ANY WARRANTY; without even the implied warranty of20
21
22
You should have received a copy of the GNU General Public License23
along with this program; if not, write to the Free Software24
-Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, 25
-MA 02110-1301, USA.26
\ No newline at end of file27
+Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.28
+29
+This program is also available under a commercial proprietary license.30
+For more information, contact us at license @ x265.com.31
x265_2.5.tar.gz/source/dynamicHDR10/SeiMetadataDictionary.cpp -> x265_2.6.tar.gz/source/dynamicHDR10/SeiMetadataDictionary.cpp
Changed
38
1
2
/**3
- * @file SeiMetadataDictionary.cpp4
- * @brief Defines the tagname for each metadata value in a JSON dynamic tone mapping file.5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,30
- * MA 02110-1301, USA.31
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.32
+ *33
+ * This program is also available under a commercial proprietary license.34
+ * For more information, contact us at license @ x265.com.35
**/36
37
#include "SeiMetadataDictionary.h"38
x265_2.5.tar.gz/source/dynamicHDR10/SeiMetadataDictionary.h -> x265_2.6.tar.gz/source/dynamicHDR10/SeiMetadataDictionary.h
Changed
38
1
2
/**3
- * @file SeiMetadataDictionary.h4
- * @brief Defines the tagname for each metadata value in a JSON dynamic tone mapping file.5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,30
- * MA 02110-1301, USA.31
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.32
+ *33
+ * This program is also available under a commercial proprietary license.34
+ * For more information, contact us at license @ x265.com.35
**/36
37
#ifndef SEIMETADATADICTIONARY_H38
x265_2.5.tar.gz/source/dynamicHDR10/api.cpp -> x265_2.6.tar.gz/source/dynamicHDR10/api.cpp
Changed
38
1
2
/**3
- * @file api.cpp4
- * @brief Implementation of hdr10plus API functions.5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, 30
- * MA 02110-1301, USA.31
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.32
+ *33
+ * This program is also available under a commercial proprietary license.34
+ * For more information, contact us at license @ x265.com.35
**/36
37
#include "hdr10plus.h"38
x265_2.5.tar.gz/source/dynamicHDR10/hdr10plus.h -> x265_2.6.tar.gz/source/dynamicHDR10/hdr10plus.h
Changed
38
1
2
/**3
- * @file hdr10plus.h4
- * @brief Definition of hdr10plus functions.5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,30
- * MA 02110-1301, USA.31
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.32
+ *33
+ * This program is also available under a commercial proprietary license.34
+ * For more information, contact us at license @ x265.com.35
**/36
37
#include <stdint.h>38
x265_2.5.tar.gz/source/dynamicHDR10/metadataFromJson.cpp -> x265_2.6.tar.gz/source/dynamicHDR10/metadataFromJson.cpp
Changed
50
1
2
/**3
- * @file metadataFromJson.cpp4
- * @brief Reads a JSON file and produces a byte array containing the metadata to be embedded in different ways.5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,30
- * MA 02110-1301, USA.31
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.32
+ *33
+ * This program is also available under a commercial proprietary license.34
+ * For more information, contact us at license @ x265.com.35
**/36
37
#include "metadataFromJson.h"38
39
40
#include "BasicStructures.h"41
#include "SeiMetadataDictionary.h"42
-43
-#define M_PI 3.1415926535897932384644
-45
using namespace SeiMetadataDictionary;46
-47
class metadataFromJson::DynamicMetaIO48
{49
public:50
x265_2.5.tar.gz/source/dynamicHDR10/metadataFromJson.h -> x265_2.6.tar.gz/source/dynamicHDR10/metadataFromJson.h
Changed
39
1
2
/**3
- * @file metadataFromJson.h4
- * @brief Reads a JSON file and produces a byte array containing the metadata to be embedded for a frame.5
- * @author Daniel Maximiliano Valenzuela, Seongnam Oh.6
- * @create date 03/01/20177
- * @version 0.0.18
+ * Copyright (C) 2013-2017 MulticoreWare, Inc9
*10
- * Copyright @ 2017 Samsung Electronics, DMS Lab, Samsung Research America and Samsung Research Tijuana11
+ * Authors: Bhavna Hariharan <bhavna@multicorewareinc.com>12
+ * Kavitha Sampath <kavitha@multicorewareinc.com>13
*14
- * This program is free software; you can redistribute it and/or15
- * modify it under the terms of the GNU General Public License16
- * as published by the Free Software Foundation; either version 217
- * of the License, or (at your option) any later version.18
+ * This program is free software; you can redistribute it and/or modify19
+ * it under the terms of the GNU General Public License as published by20
+ * the Free Software Foundation; either version 2 of the License, or21
+ * (at your option) any later version.22
*23
* This program is distributed in the hope that it will be useful,24
* but WITHOUT ANY WARRANTY; without even the implied warranty of25
26
*27
* You should have received a copy of the GNU General Public License28
* along with this program; if not, write to the Free Software29
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,30
- * MA 02110-1301, USA.31
-32
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.33
+ *34
+ * This program is also available under a commercial proprietary license.35
+ * For more information, contact us at license @ x265.com.36
**/37
38
#ifndef METADATAFROMJSON_H39
x265_2.5.tar.gz/source/encoder/CMakeLists.txt -> x265_2.6.tar.gz/source/encoder/CMakeLists.txt
Changed
8
1
2
reference.cpp reference.h3
encoder.cpp encoder.h4
api.cpp5
- weightPrediction.cpp6
- ../x265-extras.cpp ../x265-extras.h)7
+ weightPrediction.cpp)8
x265_2.5.tar.gz/source/encoder/analysis.cpp -> x265_2.6.tar.gz/source/encoder/analysis.cpp
Changed
2349
1
2
m_reuseInterDataCTU = NULL;3
m_reuseRef = NULL;4
m_bHD = false;5
+ m_modeFlag[0] = false;6
+ m_modeFlag[1] = false;7
+ m_checkMergeAndSkipOnly[0] = false;8
+ m_checkMergeAndSkipOnly[1] = false;9
m_evaluateInter = 0;10
}11
12
13
}14
else15
{16
+ bool bCopyAnalysis = ((m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel == 10) || (m_param->bMVType && m_param->analysisReuseLevel >= 7 && ctu.m_numPartitions <= 16));17
+ bool BCompressInterCUrd0_4 = (m_param->bMVType && m_param->analysisReuseLevel >= 7 && m_param->rdLevel <= 4);18
+ bool BCompressInterCUrd5_6 = (m_param->bMVType && m_param->analysisReuseLevel >= 7 && m_param->rdLevel >= 5 && m_param->rdLevel <= 6);19
+ bCopyAnalysis = bCopyAnalysis || BCompressInterCUrd0_4 || BCompressInterCUrd5_6;20
+21
+ if (bCopyAnalysis)22
+ {23
+ analysis_inter_data* interDataCTU = (analysis_inter_data*)m_frame->m_analysisData.interData;24
+ int posCTU = ctu.m_cuAddr * numPartition;25
+ memcpy(ctu.m_cuDepth, &interDataCTU->depth[posCTU], sizeof(uint8_t) * numPartition);26
+ memcpy(ctu.m_predMode, &interDataCTU->modes[posCTU], sizeof(uint8_t) * numPartition);27
+ memcpy(ctu.m_partSize, &interDataCTU->partSize[posCTU], sizeof(uint8_t) * numPartition);28
+ for (int list = 0; list < m_slice->isInterB() + 1; list++)29
+ memcpy(ctu.m_skipFlag[list], &m_frame->m_analysisData.modeFlag[list][posCTU], sizeof(uint8_t) * numPartition);30
+31
+ if ((m_slice->m_sliceType == P_SLICE || m_param->bIntraInBFrames) && !m_param->bMVType)32
+ {33
+ analysis_intra_data* intraDataCTU = (analysis_intra_data*)m_frame->m_analysisData.intraData;34
+ memcpy(ctu.m_lumaIntraDir, &intraDataCTU->modes[posCTU], sizeof(uint8_t) * numPartition);35
+ memcpy(ctu.m_chromaIntraDir, &intraDataCTU->chromaModes[posCTU], sizeof(uint8_t) * numPartition);36
+ }37
+ //Calculate log2CUSize from depth38
+ for (uint32_t i = 0; i < cuGeom.numPartitions; i++)39
+ ctu.m_log2CUSize[i] = (uint8_t)m_param->maxLog2CUSize - ctu.m_cuDepth[i];40
+ }41
+42
if (m_param->bIntraRefresh && m_slice->m_sliceType == P_SLICE &&43
ctu.m_cuPelX / m_param->maxCUSize >= frame.m_encData->m_pir.pirStartCol44
&& ctu.m_cuPelX / m_param->maxCUSize < frame.m_encData->m_pir.pirEndCol)45
46
/* generate residual for entire CTU at once and copy to reconPic */47
encodeResidue(ctu, cuGeom);48
}49
- else if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel == 10)50
+ else if ((m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel == 10) || ((m_param->bMVType == AVC_INFO) && m_param->analysisReuseLevel >= 7))51
{52
analysis_inter_data* interDataCTU = (analysis_inter_data*)m_frame->m_analysisData.interData;53
int posCTU = ctu.m_cuAddr * numPartition;54
memcpy(ctu.m_cuDepth, &interDataCTU->depth[posCTU], sizeof(uint8_t) * numPartition);55
memcpy(ctu.m_predMode, &interDataCTU->modes[posCTU], sizeof(uint8_t) * numPartition);56
memcpy(ctu.m_partSize, &interDataCTU->partSize[posCTU], sizeof(uint8_t) * numPartition);57
- if (m_slice->m_sliceType == P_SLICE || m_param->bIntraInBFrames)58
+ if ((m_slice->m_sliceType == P_SLICE || m_param->bIntraInBFrames) && !(m_param->bMVType == AVC_INFO))59
{60
analysis_intra_data* intraDataCTU = (analysis_intra_data*)m_frame->m_analysisData.intraData;61
memcpy(ctu.m_lumaIntraDir, &intraDataCTU->modes[posCTU], sizeof(uint8_t) * numPartition);62
63
mode = 2;64
else if (ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_2NxnU || ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_2NxnD || ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_nLx2N || ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_nRx2N)65
mode = 3;66
-67
if (ctu.m_predMode[puabsPartIdx + absPartIdx] == MODE_SKIP)68
{69
- ctu.m_encData->m_frameStats.cntSkipPu[depth] += (uint64_t)(1 << shift);70
- ctu.m_encData->m_frameStats.totalPu[depth] += (uint64_t)(1 << shift);71
+ ctu.m_encData->m_frameStats.cntSkipPu[depth] += 1ULL << shift;72
+ ctu.m_encData->m_frameStats.totalPu[depth] += 1ULL << shift;73
}74
else if (ctu.m_predMode[puabsPartIdx + absPartIdx] == MODE_INTRA)75
{76
77
}78
else79
{80
- ctu.m_encData->m_frameStats.cntIntraPu[depth] += (uint64_t)(1 << shift);81
- ctu.m_encData->m_frameStats.totalPu[depth] += (uint64_t)(1 << shift);82
+ ctu.m_encData->m_frameStats.cntIntraPu[depth] += 1ULL << shift;83
+ ctu.m_encData->m_frameStats.totalPu[depth] += 1ULL << shift;84
}85
}86
else if (mode == 3)87
{88
- ctu.m_encData->m_frameStats.cntAmp[depth] += (uint64_t)(1 << shift);89
- ctu.m_encData->m_frameStats.totalPu[depth] += (uint64_t)(1 << shift);90
+ ctu.m_encData->m_frameStats.cntAmp[depth] += 1ULL << shift;91
+ ctu.m_encData->m_frameStats.totalPu[depth] += 1ULL << shift;92
break;93
}94
else95
96
md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, parentCTU.m_cuAddr, cuGeom.absPartIdx);97
}98
99
-void Analysis::compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp)100
+uint64_t Analysis::compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp)101
{102
uint32_t depth = cuGeom.depth;103
ModeDepth& md = m_modeDepth[depth];104
105
Mode& mode = md.pred[0];106
md.bestMode = &mode;107
mode.cu.initSubCU(parentCTU, cuGeom, qp);108
- if (m_param->intraRefine != 2 || parentCTU.m_lumaIntraDir[cuGeom.absPartIdx] <= 1)109
+ bool reuseModes = !((m_param->intraRefine == 3) ||110
+ (m_param->intraRefine == 2 && parentCTU.m_lumaIntraDir[cuGeom.absPartIdx] > DC_IDX));111
+ if (reuseModes)112
{113
memcpy(mode.cu.m_lumaIntraDir, parentCTU.m_lumaIntraDir + cuGeom.absPartIdx, cuGeom.numPartitions);114
memcpy(mode.cu.m_chromaIntraDir, parentCTU.m_chromaIntraDir + cuGeom.absPartIdx, cuGeom.numPartitions);115
116
invalidateContexts(nextDepth);117
Entropy* nextContext = &m_rqt[depth].cur;118
int32_t nextQP = qp;119
+ uint64_t curCost = 0;120
+ int skipSplitCheck = 0;121
122
for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)123
{124
125
if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)126
nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));127
128
- compressIntraCU(parentCTU, childGeom, nextQP);129
+ if (m_param->bEnableSplitRdSkip)130
+ {131
+ curCost += compressIntraCU(parentCTU, childGeom, nextQP);132
+ if (m_modeDepth[depth].bestMode && curCost > m_modeDepth[depth].bestMode->rdCost)133
+ {134
+ skipSplitCheck = 1;135
+ break;136
+ }137
+ }138
+ else139
+ compressIntraCU(parentCTU, childGeom, nextQP);140
141
// Save best CU and pred data for this sub CU142
splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);143
144
memset(parentCTU.m_cuDepth + childGeom.absPartIdx, 0, childGeom.numPartitions);145
}146
}147
- nextContext->store(splitPred->contexts);148
- if (mightNotSplit)149
- addSplitFlagCost(*splitPred, cuGeom.depth);150
- else151
- updateModeCost(*splitPred);152
+ if (!skipSplitCheck)153
+ {154
+ nextContext->store(splitPred->contexts);155
+ if (mightNotSplit)156
+ addSplitFlagCost(*splitPred, cuGeom.depth);157
+ else158
+ updateModeCost(*splitPred);159
160
- checkDQPForSplitPred(*splitPred, cuGeom);161
- checkBestMode(*splitPred, depth);162
+ checkDQPForSplitPred(*splitPred, cuGeom);163
+ checkBestMode(*splitPred, depth);164
+ }165
}166
167
if (m_param->bEnableRdRefine && depth <= m_slice->m_pps->maxCuDQPDepth)168
169
md.bestMode->cu.copyToPic(depth);170
if (md.bestMode != &md.pred[PRED_SPLIT])171
md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, parentCTU.m_cuAddr, cuGeom.absPartIdx);172
+173
+ return md.bestMode->rdCost;174
}175
176
void Analysis::PMODE::processTasks(int workerThreadId)177
178
uint32_t depth = cuGeom.depth;179
uint32_t cuAddr = parentCTU.m_cuAddr;180
ModeDepth& md = m_modeDepth[depth];181
- md.bestMode = NULL;182
+183
184
if (m_param->searchMethod == X265_SEA)185
{186
187
}188
189
PicYuv& reconPic = *m_frame->m_reconPic;190
+ SplitData splitCUData;191
192
- bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);193
- bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);194
- uint32_t minDepth = topSkipMinDepth(parentCTU, cuGeom);195
- bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;196
- bool skipModes = false; /* Skip any remaining mode analyses at current depth */197
- bool skipRecursion = false; /* Skip recursion */198
- bool splitIntra = true;199
- bool skipRectAmp = false;200
- bool chooseMerge = false;201
- bool bCtuInfoCheck = false;202
- int sameContentRef = 0;203
+ bool bHEVCBlockAnalysis = (m_param->bMVType && cuGeom.numPartitions > 16);204
+ bool bRefineAVCAnalysis = (m_param->analysisReuseLevel == 7 && (m_modeFlag[0] || m_modeFlag[1]));205
+ bool bNooffloading = !m_param->bMVType;206
207
- if (m_evaluateInter == 1)208
+ if (bHEVCBlockAnalysis || bRefineAVCAnalysis || bNooffloading)209
{210
- skipRectAmp = !!md.bestMode;211
- mightSplit &= false;212
- minDepth = depth;213
- }214
+ md.bestMode = NULL;215
+ bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);216
+ bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);217
+ uint32_t minDepth = topSkipMinDepth(parentCTU, cuGeom);218
+ bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;219
+ bool skipModes = false; /* Skip any remaining mode analyses at current depth */220
+ bool skipRecursion = false; /* Skip recursion */221
+ bool splitIntra = true;222
+ bool skipRectAmp = false;223
+ bool chooseMerge = false;224
+ bool bCtuInfoCheck = false;225
+ int sameContentRef = 0;226
227
- if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)228
- m_maxTUDepth = loadTUDepth(cuGeom, parentCTU);229
+ if (m_evaluateInter)230
+ {231
+ if (m_param->interRefine == 2)232
+ {233
+ if (parentCTU.m_predMode[cuGeom.absPartIdx] == MODE_SKIP)234
+ skipModes = true;235
+ if (parentCTU.m_partSize[cuGeom.absPartIdx] == SIZE_2Nx2N)236
+ skipRectAmp = true;237
+ }238
+ mightSplit &= false;239
+ minDepth = depth;240
+ }241
242
- SplitData splitData[4];243
- splitData[0].initSplitCUData();244
- splitData[1].initSplitCUData();245
- splitData[2].initSplitCUData();246
- splitData[3].initSplitCUData();247
+ if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)248
+ m_maxTUDepth = loadTUDepth(cuGeom, parentCTU);249
250
- // avoid uninitialize value in below reference251
- if (m_param->limitModes)252
- {253
- md.pred[PRED_2Nx2N].bestME[0][0].mvCost = 0; // L0254
- md.pred[PRED_2Nx2N].bestME[0][1].mvCost = 0; // L1255
- md.pred[PRED_2Nx2N].sa8dCost = 0;256
- }257
+ SplitData splitData[4];258
+ splitData[0].initSplitCUData();259
+ splitData[1].initSplitCUData();260
+ splitData[2].initSplitCUData();261
+ splitData[3].initSplitCUData();262
263
- if (m_param->bCTUInfo && depth <= parentCTU.m_cuDepth[cuGeom.absPartIdx])264
- {265
- if (bDecidedDepth && m_additionalCtuInfo[cuGeom.absPartIdx])266
- sameContentRef = findSameContentRefCount(parentCTU, cuGeom);267
- if (depth < parentCTU.m_cuDepth[cuGeom.absPartIdx])268
+ // avoid uninitialize value in below reference269
+ if (m_param->limitModes)270
{271
- mightNotSplit &= bDecidedDepth;272
- bCtuInfoCheck = skipRecursion = false;273
- skipModes = true;274
+ md.pred[PRED_2Nx2N].bestME[0][0].mvCost = 0; // L0275
+ md.pred[PRED_2Nx2N].bestME[0][1].mvCost = 0; // L1276
+ md.pred[PRED_2Nx2N].sa8dCost = 0;277
}278
- else if (mightNotSplit && bDecidedDepth)279
+280
+ if (m_param->bCTUInfo && depth <= parentCTU.m_cuDepth[cuGeom.absPartIdx])281
{282
- if (m_additionalCtuInfo[cuGeom.absPartIdx])283
+ if (bDecidedDepth && m_additionalCtuInfo[cuGeom.absPartIdx])284
+ sameContentRef = findSameContentRefCount(parentCTU, cuGeom);285
+ if (depth < parentCTU.m_cuDepth[cuGeom.absPartIdx])286
+ {287
+ mightNotSplit &= bDecidedDepth;288
+ bCtuInfoCheck = skipRecursion = false;289
+ skipModes = true;290
+ }291
+ else if (mightNotSplit && bDecidedDepth)292
{293
- bCtuInfoCheck = skipRecursion = true;294
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);295
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);296
- checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);297
- if (!sameContentRef)298
+ if (m_additionalCtuInfo[cuGeom.absPartIdx])299
{300
- if ((m_param->bCTUInfo & 2) && (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth))301
+ bCtuInfoCheck = skipRecursion = true;302
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);303
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);304
+ checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);305
+ if (!sameContentRef)306
{307
- qp -= int32_t(0.04 * qp);308
- setLambdaFromQP(parentCTU, qp);309
+ if ((m_param->bCTUInfo & 2) && (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth))310
+ {311
+ qp -= int32_t(0.04 * qp);312
+ setLambdaFromQP(parentCTU, qp);313
+ }314
+ if (m_param->bCTUInfo & 4)315
+ skipModes = false;316
+ }317
+ if (sameContentRef || (!sameContentRef && !(m_param->bCTUInfo & 4)))318
+ {319
+ if (m_param->rdLevel)320
+ skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0);321
+ if ((m_param->bCTUInfo & 4) && sameContentRef)322
+ skipModes = md.bestMode && true;323
}324
- if (m_param->bCTUInfo & 4)325
- skipModes = false;326
}327
- if (sameContentRef || (!sameContentRef && !(m_param->bCTUInfo & 4)))328
+ else329
{330
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);331
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);332
+ checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);333
if (m_param->rdLevel)334
skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0);335
- if ((m_param->bCTUInfo & 4) && sameContentRef)336
- skipModes = md.bestMode && true;337
}338
+ mightSplit &= !bDecidedDepth;339
}340
- else341
- {342
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);343
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);344
- checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);345
- if (m_param->rdLevel)346
- skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0);347
- }348
- mightSplit &= !bDecidedDepth;349
}350
- }351
- if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)352
- {353
- if (mightNotSplit && depth == m_reuseDepth[cuGeom.absPartIdx])354
+ if ((m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10))355
{356
- if (m_reuseModes[cuGeom.absPartIdx] == MODE_SKIP)357
+ if (mightNotSplit && depth == m_reuseDepth[cuGeom.absPartIdx])358
{359
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);360
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);361
- checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);362
+ if (m_reuseModes[cuGeom.absPartIdx] == MODE_SKIP)363
+ {364
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);365
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);366
+ checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);367
368
- skipRecursion = !!m_param->bEnableRecursionSkip && md.bestMode;369
- if (m_param->rdLevel)370
- skipModes = m_param->bEnableEarlySkip && md.bestMode;371
- }372
- if (m_param->analysisReuseLevel > 4 && m_reusePartSize[cuGeom.absPartIdx] == SIZE_2Nx2N)373
- {374
- if (m_reuseModes[cuGeom.absPartIdx] != MODE_INTRA && m_reuseModes[cuGeom.absPartIdx] != 4)375
+ skipRecursion = !!m_param->bEnableRecursionSkip && md.bestMode;376
+ if (m_param->rdLevel)377
+ skipModes = m_param->bEnableEarlySkip && md.bestMode;378
+ }379
+ if (m_param->analysisReuseLevel > 4 && m_reusePartSize[cuGeom.absPartIdx] == SIZE_2Nx2N)380
{381
- skipRectAmp = true && !!md.bestMode;382
- chooseMerge = !!m_reuseMergeFlag[cuGeom.absPartIdx] && !!md.bestMode;383
+ if (m_reuseModes[cuGeom.absPartIdx] != MODE_INTRA && m_reuseModes[cuGeom.absPartIdx] != 4)384
+ {385
+ skipRectAmp = true && !!md.bestMode;386
+ chooseMerge = !!m_reuseMergeFlag[cuGeom.absPartIdx] && !!md.bestMode;387
+ }388
}389
}390
}391
- }392
- if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead && m_multipassAnalysis)393
- {394
- if (mightNotSplit && depth == m_multipassDepth[cuGeom.absPartIdx])395
+ if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead && m_multipassAnalysis)396
{397
- if (m_multipassModes[cuGeom.absPartIdx] == MODE_SKIP)398
+ if (mightNotSplit && depth == m_multipassDepth[cuGeom.absPartIdx])399
{400
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);401
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);402
- checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);403
+ if (m_multipassModes[cuGeom.absPartIdx] == MODE_SKIP)404
+ {405
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);406
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);407
+ checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);408
409
- skipRecursion = !!m_param->bEnableRecursionSkip && md.bestMode;410
- if (m_param->rdLevel)411
- skipModes = m_param->bEnableEarlySkip && md.bestMode;412
+ skipRecursion = !!m_param->bEnableRecursionSkip && md.bestMode;413
+ if (m_param->rdLevel)414
+ skipModes = m_param->bEnableEarlySkip && md.bestMode;415
+ }416
}417
}418
- }419
420
- /* Step 1. Evaluate Merge/Skip candidates for likely early-outs, if skip mode was not set above */421
- if (mightNotSplit && depth >= minDepth && !md.bestMode && !bCtuInfoCheck) /* TODO: Re-evaluate if analysis load/save still works */422
- {423
- /* Compute Merge Cost */424
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);425
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);426
- checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);427
- if (m_param->rdLevel)428
- skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0); // TODO: sa8d threshold per depth429
- }430
-431
- if (md.bestMode && m_param->bEnableRecursionSkip && !bCtuInfoCheck)432
- {433
- skipRecursion = md.bestMode->cu.isSkipped(0);434
- if (mightSplit && depth >= minDepth && !skipRecursion)435
+ /* Step 1. Evaluate Merge/Skip candidates for likely early-outs, if skip mode was not set above */436
+ if ((mightNotSplit && depth >= minDepth && !md.bestMode && !bCtuInfoCheck) || (m_param->bMVType && (m_modeFlag[0] || m_modeFlag[1]))) /* TODO: Re-evaluate if analysis load/save still works */437
{438
- if (depth)439
- skipRecursion = recursionDepthCheck(parentCTU, cuGeom, *md.bestMode);440
- if (m_bHD && !skipRecursion && m_param->rdLevel == 2 && md.fencYuv.m_size != MAX_CU_SIZE)441
- skipRecursion = complexityCheckCU(*md.bestMode);442
+ /* Compute Merge Cost */443
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);444
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);445
+ checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);446
+ if (m_param->rdLevel)447
+ skipModes = (m_param->bEnableEarlySkip || m_param->interRefine == 2)448
+ && md.bestMode && md.bestMode->cu.isSkipped(0); // TODO: sa8d threshold per depth449
}450
- }451
452
- /* Step 2. Evaluate each of the 4 split sub-blocks in series */453
- if (mightSplit && !skipRecursion)454
- {455
- if (bCtuInfoCheck && m_param->bCTUInfo & 2)456
- qp = int((1 / 0.96) * qp + 0.5);457
- Mode* splitPred = &md.pred[PRED_SPLIT];458
- splitPred->initCosts();459
- CUData* splitCU = &splitPred->cu;460
- splitCU->initSubCU(parentCTU, cuGeom, qp);461
+ if (md.bestMode && m_param->bEnableRecursionSkip && !bCtuInfoCheck && !(m_param->bMVType && (m_modeFlag[0] || m_modeFlag[1])))462
+ {463
+ skipRecursion = md.bestMode->cu.isSkipped(0);464
+ if (mightSplit && depth >= minDepth && !skipRecursion)465
+ {466
+ if (depth)467
+ skipRecursion = recursionDepthCheck(parentCTU, cuGeom, *md.bestMode);468
+ if (m_bHD && !skipRecursion && m_param->rdLevel == 2 && md.fencYuv.m_size != MAX_CU_SIZE)469
+ skipRecursion = complexityCheckCU(*md.bestMode);470
+ }471
+ }472
473
- uint32_t nextDepth = depth + 1;474
- ModeDepth& nd = m_modeDepth[nextDepth];475
- invalidateContexts(nextDepth);476
- Entropy* nextContext = &m_rqt[depth].cur;477
- int nextQP = qp;478
- splitIntra = false;479
+ if (m_param->bMVType && md.bestMode && cuGeom.numPartitions <= 16)480
+ skipRecursion = true;481
482
- for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)483
+ /* Step 2. Evaluate each of the 4 split sub-blocks in series */484
+ if (mightSplit && !skipRecursion)485
{486
- const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);487
- if (childGeom.flags & CUGeom::PRESENT)488
+ if (bCtuInfoCheck && m_param->bCTUInfo & 2)489
+ qp = int((1 / 0.96) * qp + 0.5);490
+ Mode* splitPred = &md.pred[PRED_SPLIT];491
+ splitPred->initCosts();492
+ CUData* splitCU = &splitPred->cu;493
+ splitCU->initSubCU(parentCTU, cuGeom, qp);494
+495
+ uint32_t nextDepth = depth + 1;496
+ ModeDepth& nd = m_modeDepth[nextDepth];497
+ invalidateContexts(nextDepth);498
+ Entropy* nextContext = &m_rqt[depth].cur;499
+ int nextQP = qp;500
+ splitIntra = false;501
+502
+ for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)503
{504
- m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);505
- m_rqt[nextDepth].cur.load(*nextContext);506
+ const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);507
+ if (childGeom.flags & CUGeom::PRESENT)508
+ {509
+ m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);510
+ m_rqt[nextDepth].cur.load(*nextContext);511
512
- if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)513
- nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));514
+ if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)515
+ nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));516
517
- splitData[subPartIdx] = compressInterCU_rd0_4(parentCTU, childGeom, nextQP);518
+ splitData[subPartIdx] = compressInterCU_rd0_4(parentCTU, childGeom, nextQP);519
520
- // Save best CU and pred data for this sub CU521
- splitIntra |= nd.bestMode->cu.isIntra(0);522
- splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);523
- splitPred->addSubCosts(*nd.bestMode);524
+ // Save best CU and pred data for this sub CU525
+ splitIntra |= nd.bestMode->cu.isIntra(0);526
+ splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);527
+ splitPred->addSubCosts(*nd.bestMode);528
529
- if (m_param->rdLevel)530
- nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);531
+ if (m_param->rdLevel)532
+ nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);533
+ else534
+ nd.bestMode->predYuv.copyToPartYuv(splitPred->predYuv, childGeom.numPartitions * subPartIdx);535
+ if (m_param->rdLevel > 1)536
+ nextContext = &nd.bestMode->contexts;537
+ }538
else539
- nd.bestMode->predYuv.copyToPartYuv(splitPred->predYuv, childGeom.numPartitions * subPartIdx);540
- if (m_param->rdLevel > 1)541
- nextContext = &nd.bestMode->contexts;542
+ splitCU->setEmptyPart(childGeom, subPartIdx);543
}544
+ nextContext->store(splitPred->contexts);545
+546
+ if (mightNotSplit)547
+ addSplitFlagCost(*splitPred, cuGeom.depth);548
+ else if (m_param->rdLevel > 1)549
+ updateModeCost(*splitPred);550
else551
- splitCU->setEmptyPart(childGeom, subPartIdx);552
+ splitPred->sa8dCost = m_rdCost.calcRdSADCost((uint32_t)splitPred->distortion, splitPred->sa8dBits);553
}554
- nextContext->store(splitPred->contexts);555
556
- if (mightNotSplit)557
- addSplitFlagCost(*splitPred, cuGeom.depth);558
- else if (m_param->rdLevel > 1)559
- updateModeCost(*splitPred);560
- else561
- splitPred->sa8dCost = m_rdCost.calcRdSADCost((uint32_t)splitPred->distortion, splitPred->sa8dBits);562
- }563
+ /* If analysis mode is simple do not Evaluate other modes */564
+ if ((m_param->bMVType && cuGeom.numPartitions <= 16) && (m_slice->m_sliceType == P_SLICE || m_slice->m_sliceType == B_SLICE))565
+ mightNotSplit = !(m_checkMergeAndSkipOnly[0] || (m_checkMergeAndSkipOnly[0] && m_checkMergeAndSkipOnly[1]));566
567
- /* Split CUs568
- * 0 1569
- * 2 3 */570
- uint32_t allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;571
- /* Step 3. Evaluate ME (2Nx2N, rect, amp) and intra modes at current depth */572
- if (mightNotSplit && (depth >= minDepth || (m_param->bCTUInfo && !md.bestMode)))573
- {574
- if (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth && m_slice->m_pps->maxCuDQPDepth != 0)575
- setLambdaFromQP(parentCTU, qp);576
-577
- if (!skipModes)578
+ /* Split CUs579
+ * 0 1580
+ * 2 3 */581
+ uint32_t allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;582
+ /* Step 3. Evaluate ME (2Nx2N, rect, amp) and intra modes at current depth */583
+ if (mightNotSplit && (depth >= minDepth || (m_param->bCTUInfo && !md.bestMode)))584
{585
- uint32_t refMasks[2];586
- refMasks[0] = allSplitRefs;587
- md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);588
- checkInter_rd0_4(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);589
-590
- if (m_param->limitReferences & X265_REF_LIMIT_CU)591
- {592
- CUData& cu = md.pred[PRED_2Nx2N].cu;593
- uint32_t refMask = cu.getBestRefIdx(0);594
- allSplitRefs = splitData[0].splitRefs = splitData[1].splitRefs = splitData[2].splitRefs = splitData[3].splitRefs = refMask;595
- }596
+ if (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth && m_slice->m_pps->maxCuDQPDepth != 0)597
+ setLambdaFromQP(parentCTU, qp);598
599
- if (m_slice->m_sliceType == B_SLICE)600
+ if (!skipModes)601
{602
- md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);603
- checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], cuGeom);604
- }605
+ uint32_t refMasks[2];606
+ refMasks[0] = allSplitRefs;607
+ md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);608
+ checkInter_rd0_4(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);609
610
- Mode *bestInter = &md.pred[PRED_2Nx2N];611
- if (!skipRectAmp)612
- {613
- if (m_param->bEnableRectInter)614
+ if (m_param->limitReferences & X265_REF_LIMIT_CU)615
{616
- uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;617
- uint32_t threshold_2NxN, threshold_Nx2N;618
-619
- if (m_slice->m_sliceType == P_SLICE)620
- {621
- threshold_2NxN = splitData[0].mvCost[0] + splitData[1].mvCost[0];622
- threshold_Nx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];623
- }624
- else625
- {626
- threshold_2NxN = (splitData[0].mvCost[0] + splitData[1].mvCost[0]627
- + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;628
- threshold_Nx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]629
- + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;630
- }631
-632
- int try_2NxN_first = threshold_2NxN < threshold_Nx2N;633
- if (try_2NxN_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxN)634
- {635
- refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */636
- refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */637
- md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);638
- checkInter_rd0_4(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);639
- if (md.pred[PRED_2NxN].sa8dCost < bestInter->sa8dCost)640
- bestInter = &md.pred[PRED_2NxN];641
- }642
-643
- if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_Nx2N)644
- {645
- refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* left */646
- refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* right */647
- md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);648
- checkInter_rd0_4(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N, refMasks);649
- if (md.pred[PRED_Nx2N].sa8dCost < bestInter->sa8dCost)650
- bestInter = &md.pred[PRED_Nx2N];651
- }652
-653
- if (!try_2NxN_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxN)654
- {655
- refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */656
- refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */657
- md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);658
- checkInter_rd0_4(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);659
- if (md.pred[PRED_2NxN].sa8dCost < bestInter->sa8dCost)660
- bestInter = &md.pred[PRED_2NxN];661
- }662
+ CUData& cu = md.pred[PRED_2Nx2N].cu;663
+ uint32_t refMask = cu.getBestRefIdx(0);664
+ allSplitRefs = splitData[0].splitRefs = splitData[1].splitRefs = splitData[2].splitRefs = splitData[3].splitRefs = refMask;665
}666
667
- if (m_slice->m_sps->maxAMPDepth > depth)668
+ if (m_slice->m_sliceType == B_SLICE)669
{670
- uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;671
- uint32_t threshold_2NxnU, threshold_2NxnD, threshold_nLx2N, threshold_nRx2N;672
-673
- if (m_slice->m_sliceType == P_SLICE)674
- {675
- threshold_2NxnU = splitData[0].mvCost[0] + splitData[1].mvCost[0];676
- threshold_2NxnD = splitData[2].mvCost[0] + splitData[3].mvCost[0];677
+ md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);678
+ checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], cuGeom);679
+ }680
681
- threshold_nLx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];682
- threshold_nRx2N = splitData[1].mvCost[0] + splitData[3].mvCost[0];683
- }684
- else685
+ Mode *bestInter = &md.pred[PRED_2Nx2N];686
+ if (!skipRectAmp)687
+ {688
+ if (m_param->bEnableRectInter)689
{690
- threshold_2NxnU = (splitData[0].mvCost[0] + splitData[1].mvCost[0]691
- + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;692
- threshold_2NxnD = (splitData[2].mvCost[0] + splitData[3].mvCost[0]693
- + splitData[2].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;694
+ uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;695
+ uint32_t threshold_2NxN, threshold_Nx2N;696
697
- threshold_nLx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]698
- + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;699
- threshold_nRx2N = (splitData[1].mvCost[0] + splitData[3].mvCost[0]700
- + splitData[1].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;701
- }702
-703
- bool bHor = false, bVer = false;704
- if (bestInter->cu.m_partSize[0] == SIZE_2NxN)705
- bHor = true;706
- else if (bestInter->cu.m_partSize[0] == SIZE_Nx2N)707
- bVer = true;708
- else if (bestInter->cu.m_partSize[0] == SIZE_2Nx2N &&709
- md.bestMode && md.bestMode->cu.getQtRootCbf(0))710
- {711
- bHor = true;712
- bVer = true;713
- }714
+ if (m_slice->m_sliceType == P_SLICE)715
+ {716
+ threshold_2NxN = splitData[0].mvCost[0] + splitData[1].mvCost[0];717
+ threshold_Nx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];718
+ }719
+ else720
+ {721
+ threshold_2NxN = (splitData[0].mvCost[0] + splitData[1].mvCost[0]722
+ + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;723
+ threshold_Nx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]724
+ + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;725
+ }726
727
- if (bHor)728
- {729
- int try_2NxnD_first = threshold_2NxnD < threshold_2NxnU;730
- if (try_2NxnD_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnD)731
+ int try_2NxN_first = threshold_2NxN < threshold_Nx2N;732
+ if (try_2NxN_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxN)733
{734
- refMasks[0] = allSplitRefs; /* 75% top */735
- refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */736
- md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);737
- checkInter_rd0_4(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);738
- if (md.pred[PRED_2NxnD].sa8dCost < bestInter->sa8dCost)739
- bestInter = &md.pred[PRED_2NxnD];740
+ refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */741
+ refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */742
+ md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);743
+ checkInter_rd0_4(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);744
+ if (md.pred[PRED_2NxN].sa8dCost < bestInter->sa8dCost)745
+ bestInter = &md.pred[PRED_2NxN];746
}747
748
- if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnU)749
+ if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_Nx2N)750
{751
- refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* 25% top */752
- refMasks[1] = allSplitRefs; /* 75% bot */753
- md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp);754
- checkInter_rd0_4(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU, refMasks);755
- if (md.pred[PRED_2NxnU].sa8dCost < bestInter->sa8dCost)756
- bestInter = &md.pred[PRED_2NxnU];757
+ refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* left */758
+ refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* right */759
+ md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);760
+ checkInter_rd0_4(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N, refMasks);761
+ if (md.pred[PRED_Nx2N].sa8dCost < bestInter->sa8dCost)762
+ bestInter = &md.pred[PRED_Nx2N];763
}764
765
- if (!try_2NxnD_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnD)766
+ if (!try_2NxN_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxN)767
{768
- refMasks[0] = allSplitRefs; /* 75% top */769
- refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */770
- md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);771
- checkInter_rd0_4(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);772
- if (md.pred[PRED_2NxnD].sa8dCost < bestInter->sa8dCost)773
- bestInter = &md.pred[PRED_2NxnD];774
+ refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */775
+ refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */776
+ md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);777
+ checkInter_rd0_4(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);778
+ if (md.pred[PRED_2NxN].sa8dCost < bestInter->sa8dCost)779
+ bestInter = &md.pred[PRED_2NxN];780
}781
}782
- if (bVer)783
+784
+ if (m_slice->m_sps->maxAMPDepth > depth)785
{786
- int try_nRx2N_first = threshold_nRx2N < threshold_nLx2N;787
- if (try_nRx2N_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nRx2N)788
+ uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;789
+ uint32_t threshold_2NxnU, threshold_2NxnD, threshold_nLx2N, threshold_nRx2N;790
+791
+ if (m_slice->m_sliceType == P_SLICE)792
+ {793
+ threshold_2NxnU = splitData[0].mvCost[0] + splitData[1].mvCost[0];794
+ threshold_2NxnD = splitData[2].mvCost[0] + splitData[3].mvCost[0];795
+796
+ threshold_nLx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];797
+ threshold_nRx2N = splitData[1].mvCost[0] + splitData[3].mvCost[0];798
+ }799
+ else800
{801
- refMasks[0] = allSplitRefs; /* 75% left */802
- refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */803
- md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);804
- checkInter_rd0_4(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);805
- if (md.pred[PRED_nRx2N].sa8dCost < bestInter->sa8dCost)806
- bestInter = &md.pred[PRED_nRx2N];807
+ threshold_2NxnU = (splitData[0].mvCost[0] + splitData[1].mvCost[0]808
+ + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;809
+ threshold_2NxnD = (splitData[2].mvCost[0] + splitData[3].mvCost[0]810
+ + splitData[2].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;811
+812
+ threshold_nLx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]813
+ + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;814
+ threshold_nRx2N = (splitData[1].mvCost[0] + splitData[3].mvCost[0]815
+ + splitData[1].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;816
}817
818
- if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nLx2N)819
+ bool bHor = false, bVer = false;820
+ if (bestInter->cu.m_partSize[0] == SIZE_2NxN)821
+ bHor = true;822
+ else if (bestInter->cu.m_partSize[0] == SIZE_Nx2N)823
+ bVer = true;824
+ else if (bestInter->cu.m_partSize[0] == SIZE_2Nx2N &&825
+ md.bestMode && md.bestMode->cu.getQtRootCbf(0))826
{827
- refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* 25% left */828
- refMasks[1] = allSplitRefs; /* 75% right */829
- md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp);830
- checkInter_rd0_4(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N, refMasks);831
- if (md.pred[PRED_nLx2N].sa8dCost < bestInter->sa8dCost)832
- bestInter = &md.pred[PRED_nLx2N];833
+ bHor = true;834
+ bVer = true;835
}836
837
- if (!try_nRx2N_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nRx2N)838
+ if (bHor)839
{840
- refMasks[0] = allSplitRefs; /* 75% left */841
- refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */842
- md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);843
- checkInter_rd0_4(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);844
- if (md.pred[PRED_nRx2N].sa8dCost < bestInter->sa8dCost)845
- bestInter = &md.pred[PRED_nRx2N];846
+ int try_2NxnD_first = threshold_2NxnD < threshold_2NxnU;847
+ if (try_2NxnD_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnD)848
+ {849
+ refMasks[0] = allSplitRefs; /* 75% top */850
+ refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */851
+ md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);852
+ checkInter_rd0_4(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);853
+ if (md.pred[PRED_2NxnD].sa8dCost < bestInter->sa8dCost)854
+ bestInter = &md.pred[PRED_2NxnD];855
+ }856
+857
+ if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnU)858
+ {859
+ refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* 25% top */860
+ refMasks[1] = allSplitRefs; /* 75% bot */861
+ md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp);862
+ checkInter_rd0_4(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU, refMasks);863
+ if (md.pred[PRED_2NxnU].sa8dCost < bestInter->sa8dCost)864
+ bestInter = &md.pred[PRED_2NxnU];865
+ }866
+867
+ if (!try_2NxnD_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnD)868
+ {869
+ refMasks[0] = allSplitRefs; /* 75% top */870
+ refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */871
+ md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);872
+ checkInter_rd0_4(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);873
+ if (md.pred[PRED_2NxnD].sa8dCost < bestInter->sa8dCost)874
+ bestInter = &md.pred[PRED_2NxnD];875
+ }876
}877
- }878
- }879
- }880
- bool bTryIntra = (m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames) && cuGeom.log2CUSize != MAX_LOG2_CU_SIZE && !((m_param->bCTUInfo & 4) && bCtuInfoCheck);881
- if (m_param->rdLevel >= 3)882
- {883
- /* Calculate RD cost of best inter option */884
- if ((!m_bChromaSa8d && (m_csp != X265_CSP_I400)) || (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)) /* When m_bChromaSa8d is enabled, chroma MC has already been done */885
- {886
- uint32_t numPU = bestInter->cu.getNumPartInter(0);887
- for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)888
- {889
- PredictionUnit pu(bestInter->cu, cuGeom, puIdx);890
- motionCompensation(bestInter->cu, pu, bestInter->predYuv, false, true);891
- }892
- }893
+ if (bVer)894
+ {895
+ int try_nRx2N_first = threshold_nRx2N < threshold_nLx2N;896
+ if (try_nRx2N_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nRx2N)897
+ {898
+ refMasks[0] = allSplitRefs; /* 75% left */899
+ refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */900
+ md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);901
+ checkInter_rd0_4(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);902
+ if (md.pred[PRED_nRx2N].sa8dCost < bestInter->sa8dCost)903
+ bestInter = &md.pred[PRED_nRx2N];904
+ }905
906
- if (!chooseMerge)907
- {908
- encodeResAndCalcRdInterCU(*bestInter, cuGeom);909
- checkBestMode(*bestInter, depth);910
+ if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nLx2N)911
+ {912
+ refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* 25% left */913
+ refMasks[1] = allSplitRefs; /* 75% right */914
+ md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp);915
+ checkInter_rd0_4(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N, refMasks);916
+ if (md.pred[PRED_nLx2N].sa8dCost < bestInter->sa8dCost)917
+ bestInter = &md.pred[PRED_nLx2N];918
+ }919
920
- /* If BIDIR is available and within 17/16 of best inter option, choose by RDO */921
- if (m_slice->m_sliceType == B_SLICE && md.pred[PRED_BIDIR].sa8dCost != MAX_INT64 &&922
- md.pred[PRED_BIDIR].sa8dCost * 16 <= bestInter->sa8dCost * 17)923
- {924
- uint32_t numPU = md.pred[PRED_BIDIR].cu.getNumPartInter(0);925
- if (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)926
- for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)927
+ if (!try_nRx2N_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nRx2N)928
{929
- PredictionUnit pu(md.pred[PRED_BIDIR].cu, cuGeom, puIdx);930
- motionCompensation(md.pred[PRED_BIDIR].cu, pu, md.pred[PRED_BIDIR].predYuv, true, true);931
+ refMasks[0] = allSplitRefs; /* 75% left */932
+ refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */933
+ md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);934
+ checkInter_rd0_4(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);935
+ if (md.pred[PRED_nRx2N].sa8dCost < bestInter->sa8dCost)936
+ bestInter = &md.pred[PRED_nRx2N];937
}938
- encodeResAndCalcRdInterCU(md.pred[PRED_BIDIR], cuGeom);939
- checkBestMode(md.pred[PRED_BIDIR], depth);940
+ }941
}942
}943
-944
- if ((bTryIntra && md.bestMode->cu.getQtRootCbf(0)) ||945
- md.bestMode->sa8dCost == MAX_INT64)946
+ bool bTryIntra = (m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames) && cuGeom.log2CUSize != MAX_LOG2_CU_SIZE && !((m_param->bCTUInfo & 4) && bCtuInfoCheck);947
+ if (m_param->rdLevel >= 3)948
{949
- if (!m_param->limitReferences || splitIntra)950
+ /* Calculate RD cost of best inter option */951
+ if ((!m_bChromaSa8d && (m_csp != X265_CSP_I400)) || (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)) /* When m_bChromaSa8d is enabled, chroma MC has already been done */952
{953
- ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);954
- md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);955
- checkIntraInInter(md.pred[PRED_INTRA], cuGeom);956
- encodeIntraInInter(md.pred[PRED_INTRA], cuGeom);957
- checkBestMode(md.pred[PRED_INTRA], depth);958
- }959
- else960
- {961
- ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);962
+ uint32_t numPU = bestInter->cu.getNumPartInter(0);963
+ for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)964
+ {965
+ PredictionUnit pu(bestInter->cu, cuGeom, puIdx);966
+ motionCompensation(bestInter->cu, pu, bestInter->predYuv, false, true);967
+ }968
}969
- }970
- }971
- else972
- {973
- /* SA8D choice between merge/skip, inter, bidir, and intra */974
- if (!md.bestMode || bestInter->sa8dCost < md.bestMode->sa8dCost)975
- md.bestMode = bestInter;976
977
- if (m_slice->m_sliceType == B_SLICE &&978
- md.pred[PRED_BIDIR].sa8dCost < md.bestMode->sa8dCost)979
- md.bestMode = &md.pred[PRED_BIDIR];980
-981
- if (bTryIntra || md.bestMode->sa8dCost == MAX_INT64)982
- {983
- if (!m_param->limitReferences || splitIntra)984
+ if (!chooseMerge)985
{986
- ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);987
- md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);988
- checkIntraInInter(md.pred[PRED_INTRA], cuGeom);989
- if (md.pred[PRED_INTRA].sa8dCost < md.bestMode->sa8dCost)990
- md.bestMode = &md.pred[PRED_INTRA];991
- }992
- else993
- {994
- ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);995
+ encodeResAndCalcRdInterCU(*bestInter, cuGeom);996
+ checkBestMode(*bestInter, depth);997
+998
+ /* If BIDIR is available and within 17/16 of best inter option, choose by RDO */999
+ if (m_slice->m_sliceType == B_SLICE && md.pred[PRED_BIDIR].sa8dCost != MAX_INT64 &&1000
+ md.pred[PRED_BIDIR].sa8dCost * 16 <= bestInter->sa8dCost * 17)1001
+ {1002
+ uint32_t numPU = md.pred[PRED_BIDIR].cu.getNumPartInter(0);1003
+ if (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)1004
+ for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)1005
+ {1006
+ PredictionUnit pu(md.pred[PRED_BIDIR].cu, cuGeom, puIdx);1007
+ motionCompensation(md.pred[PRED_BIDIR].cu, pu, md.pred[PRED_BIDIR].predYuv, true, true);1008
+ }1009
+ encodeResAndCalcRdInterCU(md.pred[PRED_BIDIR], cuGeom);1010
+ checkBestMode(md.pred[PRED_BIDIR], depth);1011
+ }1012
}1013
- }1014
1015
- /* finally code the best mode selected by SA8D costs:1016
- * RD level 2 - fully encode the best mode1017
- * RD level 1 - generate recon pixels1018
- * RD level 0 - generate chroma prediction */1019
- if (md.bestMode->cu.m_mergeFlag[0] && md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N)1020
- {1021
- /* prediction already generated for this CU, and if rd level1022
- * is not 0, it is already fully encoded */1023
- }1024
- else if (md.bestMode->cu.isInter(0))1025
- {1026
- uint32_t numPU = md.bestMode->cu.getNumPartInter(0);1027
- if (m_csp != X265_CSP_I400)1028
+ if ((bTryIntra && md.bestMode->cu.getQtRootCbf(0)) ||1029
+ md.bestMode->sa8dCost == MAX_INT64)1030
{1031
- for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)1032
+ if (!m_param->limitReferences || splitIntra)1033
{1034
- PredictionUnit pu(md.bestMode->cu, cuGeom, puIdx);1035
- motionCompensation(md.bestMode->cu, pu, md.bestMode->predYuv, false, true);1036
+ ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);1037
+ md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);1038
+ checkIntraInInter(md.pred[PRED_INTRA], cuGeom);1039
+ encodeIntraInInter(md.pred[PRED_INTRA], cuGeom);1040
+ checkBestMode(md.pred[PRED_INTRA], depth);1041
}1042
- }1043
- if (m_param->rdLevel == 2)1044
- encodeResAndCalcRdInterCU(*md.bestMode, cuGeom);1045
- else if (m_param->rdLevel == 1)1046
- {1047
- /* generate recon pixels with no rate distortion considerations */1048
- CUData& cu = md.bestMode->cu;1049
-1050
- uint32_t tuDepthRange[2];1051
- cu.getInterTUQtDepthRange(tuDepthRange, 0);1052
- m_rqt[cuGeom.depth].tmpResiYuv.subtract(*md.bestMode->fencYuv, md.bestMode->predYuv, cuGeom.log2CUSize, m_frame->m_fencPic->m_picCsp);1053
- residualTransformQuantInter(*md.bestMode, cuGeom, 0, 0, tuDepthRange);1054
- if (cu.getQtRootCbf(0))1055
- md.bestMode->reconYuv.addClip(md.bestMode->predYuv, m_rqt[cuGeom.depth].tmpResiYuv, cu.m_log2CUSize[0], m_frame->m_fencPic->m_picCsp);1056
else1057
{1058
- md.bestMode->reconYuv.copyFromYuv(md.bestMode->predYuv);1059
- if (cu.m_mergeFlag[0] && cu.m_partSize[0] == SIZE_2Nx2N)1060
- cu.setPredModeSubParts(MODE_SKIP);1061
+ ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);1062
}1063
}1064
}1065
else1066
{1067
- if (m_param->rdLevel == 2)1068
- encodeIntraInInter(*md.bestMode, cuGeom);1069
- else if (m_param->rdLevel == 1)1070
- {1071
- /* generate recon pixels with no rate distortion considerations */1072
- CUData& cu = md.bestMode->cu;1073
+ /* SA8D choice between merge/skip, inter, bidir, and intra */1074
+ if (!md.bestMode || bestInter->sa8dCost < md.bestMode->sa8dCost)1075
+ md.bestMode = bestInter;1076
1077
- uint32_t tuDepthRange[2];1078
- cu.getIntraTUQtDepthRange(tuDepthRange, 0);1079
+ if (m_slice->m_sliceType == B_SLICE &&1080
+ md.pred[PRED_BIDIR].sa8dCost < md.bestMode->sa8dCost)1081
+ md.bestMode = &md.pred[PRED_BIDIR];1082
1083
- residualTransformQuantIntra(*md.bestMode, cuGeom, 0, 0, tuDepthRange);1084
+ if (bTryIntra || md.bestMode->sa8dCost == MAX_INT64)1085
+ {1086
+ if (!m_param->limitReferences || splitIntra)1087
+ {1088
+ ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);1089
+ md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);1090
+ checkIntraInInter(md.pred[PRED_INTRA], cuGeom);1091
+ if (md.pred[PRED_INTRA].sa8dCost < md.bestMode->sa8dCost)1092
+ md.bestMode = &md.pred[PRED_INTRA];1093
+ }1094
+ else1095
+ {1096
+ ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);1097
+ }1098
+ }1099
+1100
+ /* finally code the best mode selected by SA8D costs:1101
+ * RD level 2 - fully encode the best mode1102
+ * RD level 1 - generate recon pixels1103
+ * RD level 0 - generate chroma prediction */1104
+ if (md.bestMode->cu.m_mergeFlag[0] && md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N)1105
+ {1106
+ /* prediction already generated for this CU, and if rd level1107
+ * is not 0, it is already fully encoded */1108
+ }1109
+ else if (md.bestMode->cu.isInter(0))1110
+ {1111
+ uint32_t numPU = md.bestMode->cu.getNumPartInter(0);1112
if (m_csp != X265_CSP_I400)1113
{1114
- getBestIntraModeChroma(*md.bestMode, cuGeom);1115
- residualQTIntraChroma(*md.bestMode, cuGeom, 0, 0);1116
+ for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)1117
+ {1118
+ PredictionUnit pu(md.bestMode->cu, cuGeom, puIdx);1119
+ motionCompensation(md.bestMode->cu, pu, md.bestMode->predYuv, false, true);1120
+ }1121
+ }1122
+ if (m_param->rdLevel == 2)1123
+ encodeResAndCalcRdInterCU(*md.bestMode, cuGeom);1124
+ else if (m_param->rdLevel == 1)1125
+ {1126
+ /* generate recon pixels with no rate distortion considerations */1127
+ CUData& cu = md.bestMode->cu;1128
+1129
+ uint32_t tuDepthRange[2];1130
+ cu.getInterTUQtDepthRange(tuDepthRange, 0);1131
+ m_rqt[cuGeom.depth].tmpResiYuv.subtract(*md.bestMode->fencYuv, md.bestMode->predYuv, cuGeom.log2CUSize, m_frame->m_fencPic->m_picCsp);1132
+ residualTransformQuantInter(*md.bestMode, cuGeom, 0, 0, tuDepthRange);1133
+ if (cu.getQtRootCbf(0))1134
+ md.bestMode->reconYuv.addClip(md.bestMode->predYuv, m_rqt[cuGeom.depth].tmpResiYuv, cu.m_log2CUSize[0], m_frame->m_fencPic->m_picCsp);1135
+ else1136
+ {1137
+ md.bestMode->reconYuv.copyFromYuv(md.bestMode->predYuv);1138
+ if (cu.m_mergeFlag[0] && cu.m_partSize[0] == SIZE_2Nx2N)1139
+ cu.setPredModeSubParts(MODE_SKIP);1140
+ }1141
+ }1142
+ }1143
+ else1144
+ {1145
+ if (m_param->rdLevel == 2)1146
+ encodeIntraInInter(*md.bestMode, cuGeom);1147
+ else if (m_param->rdLevel == 1)1148
+ {1149
+ /* generate recon pixels with no rate distortion considerations */1150
+ CUData& cu = md.bestMode->cu;1151
+1152
+ uint32_t tuDepthRange[2];1153
+ cu.getIntraTUQtDepthRange(tuDepthRange, 0);1154
+1155
+ residualTransformQuantIntra(*md.bestMode, cuGeom, 0, 0, tuDepthRange);1156
+ if (m_csp != X265_CSP_I400)1157
+ {1158
+ getBestIntraModeChroma(*md.bestMode, cuGeom);1159
+ residualQTIntraChroma(*md.bestMode, cuGeom, 0, 0);1160
+ }1161
+ md.bestMode->reconYuv.copyFromPicYuv(reconPic, cu.m_cuAddr, cuGeom.absPartIdx); // TODO:1162
}1163
- md.bestMode->reconYuv.copyFromPicYuv(reconPic, cu.m_cuAddr, cuGeom.absPartIdx); // TODO:1164
}1165
}1166
- }1167
- } // !earlyskip1168
+ } // !earlyskip1169
1170
- if (m_bTryLossless)1171
- tryLossless(cuGeom);1172
+ if (m_bTryLossless)1173
+ tryLossless(cuGeom);1174
1175
- if (mightSplit)1176
- addSplitFlagCost(*md.bestMode, cuGeom.depth);1177
- }1178
+ if (mightSplit)1179
+ addSplitFlagCost(*md.bestMode, cuGeom.depth);1180
+ }1181
1182
- if (mightSplit && !skipRecursion)1183
- {1184
- Mode* splitPred = &md.pred[PRED_SPLIT];1185
- if (!md.bestMode)1186
- md.bestMode = splitPred;1187
- else if (m_param->rdLevel > 1)1188
- checkBestMode(*splitPred, cuGeom.depth);1189
- else if (splitPred->sa8dCost < md.bestMode->sa8dCost)1190
- md.bestMode = splitPred;1191
+ if (mightSplit && !skipRecursion)1192
+ {1193
+ Mode* splitPred = &md.pred[PRED_SPLIT];1194
+ if (!md.bestMode)1195
+ md.bestMode = splitPred;1196
+ else if (m_param->rdLevel > 1)1197
+ checkBestMode(*splitPred, cuGeom.depth);1198
+ else if (splitPred->sa8dCost < md.bestMode->sa8dCost)1199
+ md.bestMode = splitPred;1200
1201
- checkDQPForSplitPred(*md.bestMode, cuGeom);1202
- }1203
+ checkDQPForSplitPred(*md.bestMode, cuGeom);1204
+ }1205
1206
- /* determine which motion references the parent CU should search */1207
- SplitData splitCUData;1208
- splitCUData.initSplitCUData();1209
+ /* determine which motion references the parent CU should search */1210
+ splitCUData.initSplitCUData();1211
1212
- if (m_param->limitReferences & X265_REF_LIMIT_DEPTH)1213
- {1214
- if (md.bestMode == &md.pred[PRED_SPLIT])1215
- splitCUData.splitRefs = allSplitRefs;1216
- else 1217
+ if (m_param->limitReferences & X265_REF_LIMIT_DEPTH)1218
{1219
- /* use best merge/inter mode, in case of intra use 2Nx2N inter references */1220
- CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;1221
- uint32_t numPU = cu.getNumPartInter(0);1222
- for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))1223
- splitCUData.splitRefs |= cu.getBestRefIdx(subPartIdx);1224
+ if (md.bestMode == &md.pred[PRED_SPLIT])1225
+ splitCUData.splitRefs = allSplitRefs;1226
+ else1227
+ {1228
+ /* use best merge/inter mode, in case of intra use 2Nx2N inter references */1229
+ CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;1230
+ uint32_t numPU = cu.getNumPartInter(0);1231
+ for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))1232
+ splitCUData.splitRefs |= cu.getBestRefIdx(subPartIdx);1233
+ }1234
}1235
- }1236
1237
- if (m_param->limitModes)1238
- {1239
- splitCUData.mvCost[0] = md.pred[PRED_2Nx2N].bestME[0][0].mvCost; // L01240
- splitCUData.mvCost[1] = md.pred[PRED_2Nx2N].bestME[0][1].mvCost; // L11241
- splitCUData.sa8dCost = md.pred[PRED_2Nx2N].sa8dCost;1242
- }1243
- 1244
- if (mightNotSplit && md.bestMode->cu.isSkipped(0))1245
- {1246
- FrameData& curEncData = *m_frame->m_encData;1247
- FrameData::RCStatCU& cuStat = curEncData.m_cuStat[parentCTU.m_cuAddr];1248
- uint64_t temp = cuStat.avgCost[depth] * cuStat.count[depth];1249
- cuStat.count[depth] += 1;1250
- cuStat.avgCost[depth] = (temp + md.bestMode->rdCost) / cuStat.count[depth];1251
- }1252
+ if (m_param->limitModes)1253
+ {1254
+ splitCUData.mvCost[0] = md.pred[PRED_2Nx2N].bestME[0][0].mvCost; // L01255
+ splitCUData.mvCost[1] = md.pred[PRED_2Nx2N].bestME[0][1].mvCost; // L11256
+ splitCUData.sa8dCost = md.pred[PRED_2Nx2N].sa8dCost;1257
+ }1258
1259
- /* Copy best data to encData CTU and recon */1260
- md.bestMode->cu.copyToPic(depth);1261
- if (m_param->rdLevel)1262
- md.bestMode->reconYuv.copyToPicYuv(reconPic, cuAddr, cuGeom.absPartIdx);1263
+ if (mightNotSplit && md.bestMode->cu.isSkipped(0))1264
+ {1265
+ FrameData& curEncData = *m_frame->m_encData;1266
+ FrameData::RCStatCU& cuStat = curEncData.m_cuStat[parentCTU.m_cuAddr];1267
+ uint64_t temp = cuStat.avgCost[depth] * cuStat.count[depth];1268
+ cuStat.count[depth] += 1;1269
+ cuStat.avgCost[depth] = (temp + md.bestMode->rdCost) / cuStat.count[depth];1270
+ }1271
+1272
+ /* Copy best data to encData CTU and recon */1273
+ md.bestMode->cu.copyToPic(depth);1274
+ if (m_param->rdLevel)1275
+ md.bestMode->reconYuv.copyToPicYuv(reconPic, cuAddr, cuGeom.absPartIdx);1276
1277
- if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)1278
+ if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)1279
+ {1280
+ if (mightNotSplit)1281
+ {1282
+ CUData* ctu = md.bestMode->cu.m_encData->getPicCTU(parentCTU.m_cuAddr);1283
+ int8_t maxTUDepth = -1;1284
+ for (uint32_t i = 0; i < cuGeom.numPartitions; i++)1285
+ maxTUDepth = X265_MAX(maxTUDepth, md.bestMode->cu.m_tuDepth[i]);1286
+ ctu->m_refTuDepth[cuGeom.geomRecurId] = maxTUDepth;1287
+ }1288
+ }1289
+ }1290
+ else1291
{1292
- if (mightNotSplit)1293
+ if (m_param->bMVType && cuGeom.numPartitions <= 16)1294
{1295
- CUData* ctu = md.bestMode->cu.m_encData->getPicCTU(parentCTU.m_cuAddr);1296
- int8_t maxTUDepth = -1;1297
- for (uint32_t i = 0; i < cuGeom.numPartitions; i++)1298
- maxTUDepth = X265_MAX(maxTUDepth, md.bestMode->cu.m_tuDepth[i]);1299
- ctu->m_refTuDepth[cuGeom.geomRecurId] = maxTUDepth;1300
+ qprdRefine(parentCTU, cuGeom, qp, qp);1301
+1302
+ SplitData splitData[4];1303
+ splitData[0].initSplitCUData();1304
+ splitData[1].initSplitCUData();1305
+ splitData[2].initSplitCUData();1306
+ splitData[3].initSplitCUData();1307
+1308
+ uint32_t allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;1309
+1310
+ splitCUData.initSplitCUData();1311
+1312
+ if (m_param->limitReferences & X265_REF_LIMIT_DEPTH)1313
+ {1314
+ if (md.bestMode == &md.pred[PRED_SPLIT])1315
+ splitCUData.splitRefs = allSplitRefs;1316
+ else1317
+ {1318
+ /* use best merge/inter mode, in case of intra use 2Nx2N inter references */1319
+ CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;1320
+ uint32_t numPU = cu.getNumPartInter(0);1321
+ for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))1322
+ splitCUData.splitRefs |= cu.getBestRefIdx(subPartIdx);1323
+ }1324
+ }1325
+1326
+ if (m_param->limitModes)1327
+ {1328
+ splitCUData.mvCost[0] = md.pred[PRED_2Nx2N].bestME[0][0].mvCost; // L01329
+ splitCUData.mvCost[1] = md.pred[PRED_2Nx2N].bestME[0][1].mvCost; // L11330
+ splitCUData.sa8dCost = md.pred[PRED_2Nx2N].sa8dCost;1331
+ }1332
}1333
}1334
1335
1336
m_modeDepth[depth].fencYuv.m_integral[list][i][planes] = m_frame->m_encData->m_slice->m_refFrameList[list][i]->m_encData->m_meIntegral[planes] + offset;1337
}1338
1339
- bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);1340
- bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);1341
- bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;1342
- bool skipRecursion = false;1343
- bool skipModes = false;1344
- bool splitIntra = true;1345
- bool skipRectAmp = false;1346
- bool bCtuInfoCheck = false;1347
- int sameContentRef = 0;1348
+ SplitData splitCUData;1349
1350
- if (m_evaluateInter == 1)1351
- {1352
- skipRectAmp = !!md.bestMode;1353
- mightSplit &= false;1354
- }1355
+ bool bHEVCBlockAnalysis = (m_param->bMVType && cuGeom.numPartitions > 16);1356
+ bool bRefineAVCAnalysis = (m_param->analysisReuseLevel == 7 && (m_modeFlag[0] || m_modeFlag[1]));1357
+ bool bNooffloading = !m_param->bMVType;1358
1359
- // avoid uninitialize value in below reference1360
- if (m_param->limitModes)1361
+ if (bHEVCBlockAnalysis || bRefineAVCAnalysis || bNooffloading)1362
{1363
- md.pred[PRED_2Nx2N].bestME[0][0].mvCost = 0; // L01364
- md.pred[PRED_2Nx2N].bestME[0][1].mvCost = 0; // L11365
- md.pred[PRED_2Nx2N].rdCost = 0;1366
- }1367
+ bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);1368
+ bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);1369
+ bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;1370
+ bool skipRecursion = false;1371
+ bool skipModes = false;1372
+ bool splitIntra = true;1373
+ bool skipRectAmp = false;1374
+ bool bCtuInfoCheck = false;1375
+ int sameContentRef = 0;1376
1377
- if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)1378
- m_maxTUDepth = loadTUDepth(cuGeom, parentCTU);1379
+ if (m_evaluateInter)1380
+ {1381
+ if (m_param->interRefine == 2)1382
+ {1383
+ if (parentCTU.m_predMode[cuGeom.absPartIdx] == MODE_SKIP)1384
+ skipModes = true;1385
+ if (parentCTU.m_partSize[cuGeom.absPartIdx] == SIZE_2Nx2N)1386
+ skipRectAmp = true;1387
+ }1388
+ mightSplit &= false;1389
+ }1390
1391
- SplitData splitData[4];1392
- splitData[0].initSplitCUData();1393
- splitData[1].initSplitCUData();1394
- splitData[2].initSplitCUData();1395
- splitData[3].initSplitCUData();1396
- uint32_t allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;1397
- uint32_t refMasks[2];1398
- if (m_param->bCTUInfo && depth <= parentCTU.m_cuDepth[cuGeom.absPartIdx])1399
- {1400
- if (bDecidedDepth && m_additionalCtuInfo[cuGeom.absPartIdx])1401
- sameContentRef = findSameContentRefCount(parentCTU, cuGeom);1402
- if (depth < parentCTU.m_cuDepth[cuGeom.absPartIdx])1403
+ // avoid uninitialize value in below reference1404
+ if (m_param->limitModes)1405
{1406
- mightNotSplit &= bDecidedDepth;1407
- bCtuInfoCheck = skipRecursion = false;1408
- skipModes = true;1409
+ md.pred[PRED_2Nx2N].bestME[0][0].mvCost = 0; // L01410
+ md.pred[PRED_2Nx2N].bestME[0][1].mvCost = 0; // L11411
+ md.pred[PRED_2Nx2N].rdCost = 0;1412
}1413
- else if (mightNotSplit && bDecidedDepth)1414
+1415
+ if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)1416
+ m_maxTUDepth = loadTUDepth(cuGeom, parentCTU);1417
+1418
+ SplitData splitData[4];1419
+ splitData[0].initSplitCUData();1420
+ splitData[1].initSplitCUData();1421
+ splitData[2].initSplitCUData();1422
+ splitData[3].initSplitCUData();1423
+ uint32_t allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;1424
+ uint32_t refMasks[2];1425
+ if (m_param->bCTUInfo && depth <= parentCTU.m_cuDepth[cuGeom.absPartIdx])1426
{1427
- if (m_additionalCtuInfo[cuGeom.absPartIdx])1428
+ if (bDecidedDepth && m_additionalCtuInfo[cuGeom.absPartIdx])1429
+ sameContentRef = findSameContentRefCount(parentCTU, cuGeom);1430
+ if (depth < parentCTU.m_cuDepth[cuGeom.absPartIdx])1431
{1432
- bCtuInfoCheck = skipRecursion = true;1433
- refMasks[0] = allSplitRefs;1434
- md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1435
- checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1436
- checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1437
- if (!sameContentRef)1438
+ mightNotSplit &= bDecidedDepth;1439
+ bCtuInfoCheck = skipRecursion = false;1440
+ skipModes = true;1441
+ }1442
+ else if (mightNotSplit && bDecidedDepth)1443
+ {1444
+ if (m_additionalCtuInfo[cuGeom.absPartIdx])1445
{1446
- if ((m_param->bCTUInfo & 2) && (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth))1447
+ bCtuInfoCheck = skipRecursion = true;1448
+ refMasks[0] = allSplitRefs;1449
+ md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1450
+ checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1451
+ checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1452
+ if (!sameContentRef)1453
{1454
- qp -= int32_t(0.04 * qp);1455
- setLambdaFromQP(parentCTU, qp);1456
+ if ((m_param->bCTUInfo & 2) && (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth))1457
+ {1458
+ qp -= int32_t(0.04 * qp);1459
+ setLambdaFromQP(parentCTU, qp);1460
+ }1461
+ if (m_param->bCTUInfo & 4)1462
+ skipModes = false;1463
+ }1464
+ if (sameContentRef || (!sameContentRef && !(m_param->bCTUInfo & 4)))1465
+ {1466
+ if (m_param->rdLevel)1467
+ skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0);1468
+ if ((m_param->bCTUInfo & 4) && sameContentRef)1469
+ skipModes = md.bestMode && true;1470
}1471
- if (m_param->bCTUInfo & 4)1472
- skipModes = false;1473
}1474
- if (sameContentRef || (!sameContentRef && !(m_param->bCTUInfo & 4)))1475
+ else1476
{1477
- if (m_param->rdLevel)1478
- skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0);1479
- if ((m_param->bCTUInfo & 4) && sameContentRef)1480
- skipModes = md.bestMode && true;1481
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);1482
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);1483
+ checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);1484
+ skipModes = !!m_param->bEnableEarlySkip && md.bestMode;1485
+ refMasks[0] = allSplitRefs;1486
+ md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1487
+ checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1488
+ checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1489
}1490
+ mightSplit &= !bDecidedDepth;1491
}1492
- else1493
- {1494
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);1495
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);1496
- checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);1497
- skipModes = !!m_param->bEnableEarlySkip && md.bestMode;1498
- refMasks[0] = allSplitRefs;1499
- md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1500
- checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1501
- checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1502
- }1503
- mightSplit &= !bDecidedDepth;1504
}1505
- }1506
- if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)1507
- {1508
- if (mightNotSplit && depth == m_reuseDepth[cuGeom.absPartIdx])1509
+ if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)1510
{1511
- if (m_reuseModes[cuGeom.absPartIdx] == MODE_SKIP)1512
+ if (mightNotSplit && depth == m_reuseDepth[cuGeom.absPartIdx])1513
{1514
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);1515
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);1516
- checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);1517
- skipModes = !!m_param->bEnableEarlySkip && md.bestMode;1518
- refMasks[0] = allSplitRefs;1519
- md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1520
- checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1521
- checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1522
+ if (m_reuseModes[cuGeom.absPartIdx] == MODE_SKIP)1523
+ {1524
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);1525
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);1526
+ checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);1527
+ skipModes = !!m_param->bEnableEarlySkip && md.bestMode;1528
+ refMasks[0] = allSplitRefs;1529
+ md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1530
+ checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1531
+ checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1532
1533
- if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)1534
- skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);1535
+ if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)1536
+ skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);1537
+ }1538
+ if (m_param->analysisReuseLevel > 4 && m_reusePartSize[cuGeom.absPartIdx] == SIZE_2Nx2N)1539
+ skipRectAmp = true && !!md.bestMode;1540
}1541
- if (m_param->analysisReuseLevel > 4 && m_reusePartSize[cuGeom.absPartIdx] == SIZE_2Nx2N)1542
- skipRectAmp = true && !!md.bestMode;1543
}1544
- }1545
1546
- if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead && m_multipassAnalysis)1547
- {1548
- if (mightNotSplit && depth == m_multipassDepth[cuGeom.absPartIdx])1549
+ if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead && m_multipassAnalysis)1550
{1551
- if (m_multipassModes[cuGeom.absPartIdx] == MODE_SKIP)1552
+ if (mightNotSplit && depth == m_multipassDepth[cuGeom.absPartIdx])1553
{1554
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);1555
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);1556
- checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);1557
+ if (m_multipassModes[cuGeom.absPartIdx] == MODE_SKIP)1558
+ {1559
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);1560
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);1561
+ checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);1562
1563
- skipModes = !!m_param->bEnableEarlySkip && md.bestMode;1564
- refMasks[0] = allSplitRefs;1565
- md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1566
- checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1567
- checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1568
+ skipModes = !!m_param->bEnableEarlySkip && md.bestMode;1569
+ refMasks[0] = allSplitRefs;1570
+ md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1571
+ checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1572
+ checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1573
1574
- if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)1575
- skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);1576
+ if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)1577
+ skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);1578
+ }1579
}1580
}1581
- }1582
1583
- /* Step 1. Evaluate Merge/Skip candidates for likely early-outs */1584
- if (mightNotSplit && !md.bestMode && !bCtuInfoCheck)1585
- {1586
- md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);1587
- md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);1588
- checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);1589
- skipModes = m_param->bEnableEarlySkip && md.bestMode && !md.bestMode->cu.getQtRootCbf(0);1590
- refMasks[0] = allSplitRefs;1591
- md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1592
- checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1593
- checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1594
-1595
- if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)1596
- skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);1597
- }1598
-1599
- // estimate split cost1600
- /* Step 2. Evaluate each of the 4 split sub-blocks in series */1601
- if (mightSplit && !skipRecursion)1602
- {1603
- if (bCtuInfoCheck && m_param->bCTUInfo & 2)1604
- qp = int((1 / 0.96) * qp + 0.5);1605
- Mode* splitPred = &md.pred[PRED_SPLIT];1606
- splitPred->initCosts();1607
- CUData* splitCU = &splitPred->cu;1608
- splitCU->initSubCU(parentCTU, cuGeom, qp);1609
+ /* Step 1. Evaluate Merge/Skip candidates for likely early-outs */1610
+ if ((mightNotSplit && !md.bestMode && !bCtuInfoCheck) ||1611
+ (m_param->bMVType && (m_modeFlag[0] || m_modeFlag[1])))1612
+ {1613
+ md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);1614
+ md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);1615
+ checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);1616
+ skipModes = (m_param->bEnableEarlySkip || m_param->interRefine == 2) &&1617
+ md.bestMode && !md.bestMode->cu.getQtRootCbf(0);1618
+ refMasks[0] = allSplitRefs;1619
+ md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1620
+ checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);1621
+ checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);1622
1623
- uint32_t nextDepth = depth + 1;1624
- ModeDepth& nd = m_modeDepth[nextDepth];1625
- invalidateContexts(nextDepth);1626
- Entropy* nextContext = &m_rqt[depth].cur;1627
- int nextQP = qp;1628
- splitIntra = false;1629
+ if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)1630
+ skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);1631
+ }1632
1633
- for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)1634
+ if (m_param->bMVType && md.bestMode && cuGeom.numPartitions <= 16)1635
+ skipRecursion = true;1636
+1637
+ // estimate split cost1638
+ /* Step 2. Evaluate each of the 4 split sub-blocks in series */1639
+ if (mightSplit && !skipRecursion)1640
{1641
- const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);1642
- if (childGeom.flags & CUGeom::PRESENT)1643
+ if (bCtuInfoCheck && m_param->bCTUInfo & 2)1644
+ qp = int((1 / 0.96) * qp + 0.5);1645
+ Mode* splitPred = &md.pred[PRED_SPLIT];1646
+ splitPred->initCosts();1647
+ CUData* splitCU = &splitPred->cu;1648
+ splitCU->initSubCU(parentCTU, cuGeom, qp);1649
+1650
+ uint32_t nextDepth = depth + 1;1651
+ ModeDepth& nd = m_modeDepth[nextDepth];1652
+ invalidateContexts(nextDepth);1653
+ Entropy* nextContext = &m_rqt[depth].cur;1654
+ int nextQP = qp;1655
+ splitIntra = false;1656
+1657
+ for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)1658
{1659
- m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);1660
- m_rqt[nextDepth].cur.load(*nextContext);1661
+ const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);1662
+ if (childGeom.flags & CUGeom::PRESENT)1663
+ {1664
+ m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);1665
+ m_rqt[nextDepth].cur.load(*nextContext);1666
1667
- if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)1668
- nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));1669
+ if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)1670
+ nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));1671
1672
- splitData[subPartIdx] = compressInterCU_rd5_6(parentCTU, childGeom, nextQP);1673
+ splitData[subPartIdx] = compressInterCU_rd5_6(parentCTU, childGeom, nextQP);1674
1675
- // Save best CU and pred data for this sub CU1676
- splitIntra |= nd.bestMode->cu.isIntra(0);1677
- splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);1678
- splitPred->addSubCosts(*nd.bestMode);1679
- nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);1680
- nextContext = &nd.bestMode->contexts;1681
+ // Save best CU and pred data for this sub CU1682
+ splitIntra |= nd.bestMode->cu.isIntra(0);1683
+ splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);1684
+ splitPred->addSubCosts(*nd.bestMode);1685
+ nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);1686
+ nextContext = &nd.bestMode->contexts;1687
+ }1688
+ else1689
+ {1690
+ splitCU->setEmptyPart(childGeom, subPartIdx);1691
+ }1692
}1693
+ nextContext->store(splitPred->contexts);1694
+ if (mightNotSplit)1695
+ addSplitFlagCost(*splitPred, cuGeom.depth);1696
else1697
- {1698
- splitCU->setEmptyPart(childGeom, subPartIdx);1699
- }1700
- }1701
- nextContext->store(splitPred->contexts);1702
- if (mightNotSplit)1703
- addSplitFlagCost(*splitPred, cuGeom.depth);1704
- else1705
- updateModeCost(*splitPred);1706
+ updateModeCost(*splitPred);1707
1708
- checkDQPForSplitPred(*splitPred, cuGeom);1709
- }1710
+ checkDQPForSplitPred(*splitPred, cuGeom);1711
+ }1712
1713
- /* Split CUs1714
- * 0 11715
- * 2 3 */1716
- allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;1717
- /* Step 3. Evaluate ME (2Nx2N, rect, amp) and intra modes at current depth */1718
- if (mightNotSplit)1719
- {1720
- if (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth && m_slice->m_pps->maxCuDQPDepth != 0)1721
- setLambdaFromQP(parentCTU, qp);1722
+ /* If analysis mode is simple do not Evaluate other modes */1723
+ if ((m_param->bMVType && cuGeom.numPartitions <= 16) && (m_slice->m_sliceType == P_SLICE || m_slice->m_sliceType == B_SLICE))1724
+ mightNotSplit = !(m_checkMergeAndSkipOnly[0] || (m_checkMergeAndSkipOnly[0] && m_checkMergeAndSkipOnly[1]));1725
1726
- if (!skipModes)1727
+ /* Split CUs1728
+ * 0 11729
+ * 2 3 */1730
+ allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;1731
+ /* Step 3. Evaluate ME (2Nx2N, rect, amp) and intra modes at current depth */1732
+ if (mightNotSplit)1733
{1734
- refMasks[0] = allSplitRefs;1735
+ if (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth && m_slice->m_pps->maxCuDQPDepth != 0)1736
+ setLambdaFromQP(parentCTU, qp);1737
1738
- if (m_param->limitReferences & X265_REF_LIMIT_CU)1739
+ if (!skipModes)1740
{1741
- CUData& cu = md.pred[PRED_2Nx2N].cu;1742
- uint32_t refMask = cu.getBestRefIdx(0);1743
- allSplitRefs = splitData[0].splitRefs = splitData[1].splitRefs = splitData[2].splitRefs = splitData[3].splitRefs = refMask;1744
- }1745
+ refMasks[0] = allSplitRefs;1746
1747
- if (m_slice->m_sliceType == B_SLICE)1748
- {1749
- md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);1750
- checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], cuGeom);1751
- if (md.pred[PRED_BIDIR].sa8dCost < MAX_INT64)1752
+ if (m_param->limitReferences & X265_REF_LIMIT_CU)1753
{1754
- uint32_t numPU = md.pred[PRED_BIDIR].cu.getNumPartInter(0);1755
- if (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)1756
- for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)1757
- {1758
- PredictionUnit pu(md.pred[PRED_BIDIR].cu, cuGeom, puIdx);1759
- motionCompensation(md.pred[PRED_BIDIR].cu, pu, md.pred[PRED_BIDIR].predYuv, true, true);1760
- }1761
- encodeResAndCalcRdInterCU(md.pred[PRED_BIDIR], cuGeom);1762
- checkBestMode(md.pred[PRED_BIDIR], cuGeom.depth);1763
+ CUData& cu = md.pred[PRED_2Nx2N].cu;1764
+ uint32_t refMask = cu.getBestRefIdx(0);1765
+ allSplitRefs = splitData[0].splitRefs = splitData[1].splitRefs = splitData[2].splitRefs = splitData[3].splitRefs = refMask;1766
}1767
- }1768
1769
- if (!skipRectAmp)1770
- {1771
- if (m_param->bEnableRectInter)1772
+ if (m_slice->m_sliceType == B_SLICE)1773
{1774
- uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;1775
- uint32_t threshold_2NxN, threshold_Nx2N;1776
-1777
- if (m_slice->m_sliceType == P_SLICE)1778
- {1779
- threshold_2NxN = splitData[0].mvCost[0] + splitData[1].mvCost[0];1780
- threshold_Nx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];1781
- }1782
- else1783
- {1784
- threshold_2NxN = (splitData[0].mvCost[0] + splitData[1].mvCost[0]1785
- + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;1786
- threshold_Nx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]1787
- + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;1788
- }1789
-1790
- int try_2NxN_first = threshold_2NxN < threshold_Nx2N;1791
- if (try_2NxN_first && splitCost < md.bestMode->rdCost + threshold_2NxN)1792
- {1793
- refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */1794
- refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */1795
- md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);1796
- checkInter_rd5_6(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);1797
- checkBestMode(md.pred[PRED_2NxN], cuGeom.depth);1798
- }1799
-1800
- if (splitCost < md.bestMode->rdCost + threshold_Nx2N)1801
- {1802
- refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* left */1803
- refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* right */1804
- md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1805
- checkInter_rd5_6(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N, refMasks);1806
- checkBestMode(md.pred[PRED_Nx2N], cuGeom.depth);1807
- }1808
-1809
- if (!try_2NxN_first && splitCost < md.bestMode->rdCost + threshold_2NxN)1810
+ md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);1811
+ checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], cuGeom);1812
+ if (md.pred[PRED_BIDIR].sa8dCost < MAX_INT64)1813
{1814
- refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */1815
- refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */1816
- md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);1817
- checkInter_rd5_6(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);1818
- checkBestMode(md.pred[PRED_2NxN], cuGeom.depth);1819
+ uint32_t numPU = md.pred[PRED_BIDIR].cu.getNumPartInter(0);1820
+ if (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)1821
+ for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)1822
+ {1823
+ PredictionUnit pu(md.pred[PRED_BIDIR].cu, cuGeom, puIdx);1824
+ motionCompensation(md.pred[PRED_BIDIR].cu, pu, md.pred[PRED_BIDIR].predYuv, true, true);1825
+ }1826
+ encodeResAndCalcRdInterCU(md.pred[PRED_BIDIR], cuGeom);1827
+ checkBestMode(md.pred[PRED_BIDIR], cuGeom.depth);1828
}1829
}1830
1831
- // Try AMP (SIZE_2NxnU, SIZE_2NxnD, SIZE_nLx2N, SIZE_nRx2N)1832
- if (m_slice->m_sps->maxAMPDepth > depth)1833
+ if (!skipRectAmp)1834
{1835
- uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;1836
- uint32_t threshold_2NxnU, threshold_2NxnD, threshold_nLx2N, threshold_nRx2N;1837
-1838
- if (m_slice->m_sliceType == P_SLICE)1839
+ if (m_param->bEnableRectInter)1840
{1841
- threshold_2NxnU = splitData[0].mvCost[0] + splitData[1].mvCost[0];1842
- threshold_2NxnD = splitData[2].mvCost[0] + splitData[3].mvCost[0];1843
+ uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;1844
+ uint32_t threshold_2NxN, threshold_Nx2N;1845
1846
- threshold_nLx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];1847
- threshold_nRx2N = splitData[1].mvCost[0] + splitData[3].mvCost[0];1848
- }1849
- else1850
- {1851
- threshold_2NxnU = (splitData[0].mvCost[0] + splitData[1].mvCost[0]1852
- + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;1853
- threshold_2NxnD = (splitData[2].mvCost[0] + splitData[3].mvCost[0]1854
- + splitData[2].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;1855
-1856
- threshold_nLx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]1857
- + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;1858
- threshold_nRx2N = (splitData[1].mvCost[0] + splitData[3].mvCost[0]1859
- + splitData[1].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;1860
- }1861
-1862
- bool bHor = false, bVer = false;1863
- if (md.bestMode->cu.m_partSize[0] == SIZE_2NxN)1864
- bHor = true;1865
- else if (md.bestMode->cu.m_partSize[0] == SIZE_Nx2N)1866
- bVer = true;1867
- else if (md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N && !md.bestMode->cu.m_mergeFlag[0])1868
- {1869
- bHor = true;1870
- bVer = true;1871
- }1872
+ if (m_slice->m_sliceType == P_SLICE)1873
+ {1874
+ threshold_2NxN = splitData[0].mvCost[0] + splitData[1].mvCost[0];1875
+ threshold_Nx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];1876
+ }1877
+ else1878
+ {1879
+ threshold_2NxN = (splitData[0].mvCost[0] + splitData[1].mvCost[0]1880
+ + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;1881
+ threshold_Nx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]1882
+ + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;1883
+ }1884
1885
- if (bHor)1886
- {1887
- int try_2NxnD_first = threshold_2NxnD < threshold_2NxnU;1888
- if (try_2NxnD_first && splitCost < md.bestMode->rdCost + threshold_2NxnD)1889
+ int try_2NxN_first = threshold_2NxN < threshold_Nx2N;1890
+ if (try_2NxN_first && splitCost < md.bestMode->rdCost + threshold_2NxN)1891
{1892
- refMasks[0] = allSplitRefs; /* 75% top */1893
- refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */1894
- md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);1895
- checkInter_rd5_6(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);1896
- checkBestMode(md.pred[PRED_2NxnD], cuGeom.depth);1897
+ refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */1898
+ refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */1899
+ md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);1900
+ checkInter_rd5_6(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);1901
+ checkBestMode(md.pred[PRED_2NxN], cuGeom.depth);1902
}1903
1904
- if (splitCost < md.bestMode->rdCost + threshold_2NxnU)1905
+ if (splitCost < md.bestMode->rdCost + threshold_Nx2N)1906
{1907
- refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* 25% top */1908
- refMasks[1] = allSplitRefs; /* 75% bot */1909
- md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp);1910
- checkInter_rd5_6(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU, refMasks);1911
- checkBestMode(md.pred[PRED_2NxnU], cuGeom.depth);1912
+ refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* left */1913
+ refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* right */1914
+ md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);1915
+ checkInter_rd5_6(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N, refMasks);1916
+ checkBestMode(md.pred[PRED_Nx2N], cuGeom.depth);1917
}1918
1919
- if (!try_2NxnD_first && splitCost < md.bestMode->rdCost + threshold_2NxnD)1920
+ if (!try_2NxN_first && splitCost < md.bestMode->rdCost + threshold_2NxN)1921
{1922
- refMasks[0] = allSplitRefs; /* 75% top */1923
- refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */1924
- md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);1925
- checkInter_rd5_6(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);1926
- checkBestMode(md.pred[PRED_2NxnD], cuGeom.depth);1927
+ refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */1928
+ refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */1929
+ md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);1930
+ checkInter_rd5_6(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);1931
+ checkBestMode(md.pred[PRED_2NxN], cuGeom.depth);1932
}1933
}1934
1935
- if (bVer)1936
+ // Try AMP (SIZE_2NxnU, SIZE_2NxnD, SIZE_nLx2N, SIZE_nRx2N)1937
+ if (m_slice->m_sps->maxAMPDepth > depth)1938
{1939
- int try_nRx2N_first = threshold_nRx2N < threshold_nLx2N;1940
- if (try_nRx2N_first && splitCost < md.bestMode->rdCost + threshold_nRx2N)1941
+ uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;1942
+ uint32_t threshold_2NxnU, threshold_2NxnD, threshold_nLx2N, threshold_nRx2N;1943
+1944
+ if (m_slice->m_sliceType == P_SLICE)1945
+ {1946
+ threshold_2NxnU = splitData[0].mvCost[0] + splitData[1].mvCost[0];1947
+ threshold_2NxnD = splitData[2].mvCost[0] + splitData[3].mvCost[0];1948
+1949
+ threshold_nLx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];1950
+ threshold_nRx2N = splitData[1].mvCost[0] + splitData[3].mvCost[0];1951
+ }1952
+ else1953
{1954
- refMasks[0] = allSplitRefs; /* 75% left */1955
- refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */1956
- md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);1957
- checkInter_rd5_6(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);1958
- checkBestMode(md.pred[PRED_nRx2N], cuGeom.depth);1959
+ threshold_2NxnU = (splitData[0].mvCost[0] + splitData[1].mvCost[0]1960
+ + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;1961
+ threshold_2NxnD = (splitData[2].mvCost[0] + splitData[3].mvCost[0]1962
+ + splitData[2].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;1963
+1964
+ threshold_nLx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]1965
+ + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;1966
+ threshold_nRx2N = (splitData[1].mvCost[0] + splitData[3].mvCost[0]1967
+ + splitData[1].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;1968
}1969
1970
- if (splitCost < md.bestMode->rdCost + threshold_nLx2N)1971
+ bool bHor = false, bVer = false;1972
+ if (md.bestMode->cu.m_partSize[0] == SIZE_2NxN)1973
+ bHor = true;1974
+ else if (md.bestMode->cu.m_partSize[0] == SIZE_Nx2N)1975
+ bVer = true;1976
+ else if (md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N && !md.bestMode->cu.m_mergeFlag[0])1977
{1978
- refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* 25% left */1979
- refMasks[1] = allSplitRefs; /* 75% right */1980
- md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp);1981
- checkInter_rd5_6(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N, refMasks);1982
- checkBestMode(md.pred[PRED_nLx2N], cuGeom.depth);1983
+ bHor = true;1984
+ bVer = true;1985
}1986
1987
- if (!try_nRx2N_first && splitCost < md.bestMode->rdCost + threshold_nRx2N)1988
+ if (bHor)1989
{1990
- refMasks[0] = allSplitRefs; /* 75% left */1991
- refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */1992
- md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);1993
- checkInter_rd5_6(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);1994
- checkBestMode(md.pred[PRED_nRx2N], cuGeom.depth);1995
+ int try_2NxnD_first = threshold_2NxnD < threshold_2NxnU;1996
+ if (try_2NxnD_first && splitCost < md.bestMode->rdCost + threshold_2NxnD)1997
+ {1998
+ refMasks[0] = allSplitRefs; /* 75% top */1999
+ refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */2000
+ md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);2001
+ checkInter_rd5_6(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);2002
+ checkBestMode(md.pred[PRED_2NxnD], cuGeom.depth);2003
+ }2004
+2005
+ if (splitCost < md.bestMode->rdCost + threshold_2NxnU)2006
+ {2007
+ refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* 25% top */2008
+ refMasks[1] = allSplitRefs; /* 75% bot */2009
+ md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp);2010
+ checkInter_rd5_6(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU, refMasks);2011
+ checkBestMode(md.pred[PRED_2NxnU], cuGeom.depth);2012
+ }2013
+2014
+ if (!try_2NxnD_first && splitCost < md.bestMode->rdCost + threshold_2NxnD)2015
+ {2016
+ refMasks[0] = allSplitRefs; /* 75% top */2017
+ refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */2018
+ md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);2019
+ checkInter_rd5_6(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);2020
+ checkBestMode(md.pred[PRED_2NxnD], cuGeom.depth);2021
+ }2022
+ }2023
+2024
+ if (bVer)2025
+ {2026
+ int try_nRx2N_first = threshold_nRx2N < threshold_nLx2N;2027
+ if (try_nRx2N_first && splitCost < md.bestMode->rdCost + threshold_nRx2N)2028
+ {2029
+ refMasks[0] = allSplitRefs; /* 75% left */2030
+ refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */2031
+ md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);2032
+ checkInter_rd5_6(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);2033
+ checkBestMode(md.pred[PRED_nRx2N], cuGeom.depth);2034
+ }2035
+2036
+ if (splitCost < md.bestMode->rdCost + threshold_nLx2N)2037
+ {2038
+ refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* 25% left */2039
+ refMasks[1] = allSplitRefs; /* 75% right */2040
+ md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp);2041
+ checkInter_rd5_6(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N, refMasks);2042
+ checkBestMode(md.pred[PRED_nLx2N], cuGeom.depth);2043
+ }2044
+2045
+ if (!try_nRx2N_first && splitCost < md.bestMode->rdCost + threshold_nRx2N)2046
+ {2047
+ refMasks[0] = allSplitRefs; /* 75% left */2048
+ refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */2049
+ md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);2050
+ checkInter_rd5_6(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);2051
+ checkBestMode(md.pred[PRED_nRx2N], cuGeom.depth);2052
+ }2053
}2054
}2055
}2056
- }2057
2058
- if ((m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames) && (cuGeom.log2CUSize != MAX_LOG2_CU_SIZE) && !((m_param->bCTUInfo & 4) && bCtuInfoCheck))2059
- {2060
- if (!m_param->limitReferences || splitIntra)2061
+ if ((m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames) && (cuGeom.log2CUSize != MAX_LOG2_CU_SIZE) && !((m_param->bCTUInfo & 4) && bCtuInfoCheck))2062
{2063
- ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);2064
- md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);2065
- checkIntra(md.pred[PRED_INTRA], cuGeom, SIZE_2Nx2N);2066
- checkBestMode(md.pred[PRED_INTRA], depth);2067
+ if (!m_param->limitReferences || splitIntra)2068
+ {2069
+ ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);2070
+ md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);2071
+ checkIntra(md.pred[PRED_INTRA], cuGeom, SIZE_2Nx2N);2072
+ checkBestMode(md.pred[PRED_INTRA], depth);2073
2074
- if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3)2075
+ if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3)2076
+ {2077
+ md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom, qp);2078
+ checkIntra(md.pred[PRED_INTRA_NxN], cuGeom, SIZE_NxN);2079
+ checkBestMode(md.pred[PRED_INTRA_NxN], depth);2080
+ }2081
+ }2082
+ else2083
{2084
- md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom, qp);2085
- checkIntra(md.pred[PRED_INTRA_NxN], cuGeom, SIZE_NxN);2086
- checkBestMode(md.pred[PRED_INTRA_NxN], depth);2087
+ ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);2088
}2089
}2090
- else2091
+ }2092
+2093
+ if ((md.bestMode->cu.isInter(0) && !(md.bestMode->cu.m_mergeFlag[0] && md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N)) && (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400))2094
+ {2095
+ uint32_t numPU = md.bestMode->cu.getNumPartInter(0);2096
+2097
+ for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)2098
{2099
- ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);2100
+ PredictionUnit pu(md.bestMode->cu, cuGeom, puIdx);2101
+ motionCompensation(md.bestMode->cu, pu, md.bestMode->predYuv, false, m_csp != X265_CSP_I400);2102
}2103
+ encodeResAndCalcRdInterCU(*md.bestMode, cuGeom);2104
}2105
+ if (m_bTryLossless)2106
+ tryLossless(cuGeom);2107
+2108
+ if (mightSplit)2109
+ addSplitFlagCost(*md.bestMode, cuGeom.depth);2110
}2111
2112
- if ((md.bestMode->cu.isInter(0) && !(md.bestMode->cu.m_mergeFlag[0] && md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N)) && (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400))2113
+ if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)2114
{2115
- uint32_t numPU = md.bestMode->cu.getNumPartInter(0);2116
-2117
- for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)2118
+ if (mightNotSplit)2119
{2120
- PredictionUnit pu(md.bestMode->cu, cuGeom, puIdx);2121
- motionCompensation(md.bestMode->cu, pu, md.bestMode->predYuv, false, m_csp != X265_CSP_I400);2122
+ CUData* ctu = md.bestMode->cu.m_encData->getPicCTU(parentCTU.m_cuAddr);2123
+ int8_t maxTUDepth = -1;2124
+ for (uint32_t i = 0; i < cuGeom.numPartitions; i++)2125
+ maxTUDepth = X265_MAX(maxTUDepth, md.bestMode->cu.m_tuDepth[i]);2126
+ ctu->m_refTuDepth[cuGeom.geomRecurId] = maxTUDepth;2127
}2128
- encodeResAndCalcRdInterCU(*md.bestMode, cuGeom);2129
}2130
- if (m_bTryLossless)2131
- tryLossless(cuGeom);2132
2133
- if (mightSplit)2134
- addSplitFlagCost(*md.bestMode, cuGeom.depth);2135
- }2136
+ /* compare split RD cost against best cost */2137
+ if (mightSplit && !skipRecursion)2138
+ checkBestMode(md.pred[PRED_SPLIT], depth);2139
2140
- if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)2141
- {2142
- if (mightNotSplit)2143
+ if (m_param->bEnableRdRefine && depth <= m_slice->m_pps->maxCuDQPDepth)2144
{2145
- CUData* ctu = md.bestMode->cu.m_encData->getPicCTU(parentCTU.m_cuAddr);2146
- int8_t maxTUDepth = -1;2147
- for (uint32_t i = 0; i < cuGeom.numPartitions; i++)2148
- maxTUDepth = X265_MAX(maxTUDepth, md.bestMode->cu.m_tuDepth[i]);2149
- ctu->m_refTuDepth[cuGeom.geomRecurId] = maxTUDepth;2150
+ int cuIdx = (cuGeom.childOffset - 1) / 3;2151
+ cacheCost[cuIdx] = md.bestMode->rdCost;2152
}2153
- }2154
2155
- /* compare split RD cost against best cost */2156
- if (mightSplit && !skipRecursion)2157
- checkBestMode(md.pred[PRED_SPLIT], depth);2158
-2159
- if (m_param->bEnableRdRefine && depth <= m_slice->m_pps->maxCuDQPDepth)2160
- {2161
- int cuIdx = (cuGeom.childOffset - 1) / 3;2162
- cacheCost[cuIdx] = md.bestMode->rdCost;2163
- }2164
+ /* determine which motion references the parent CU should search */2165
+ splitCUData.initSplitCUData();2166
+ if (m_param->limitReferences & X265_REF_LIMIT_DEPTH)2167
+ {2168
+ if (md.bestMode == &md.pred[PRED_SPLIT])2169
+ splitCUData.splitRefs = allSplitRefs;2170
+ else2171
+ {2172
+ /* use best merge/inter mode, in case of intra use 2Nx2N inter references */2173
+ CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;2174
+ uint32_t numPU = cu.getNumPartInter(0);2175
+ for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))2176
+ splitCUData.splitRefs |= cu.getBestRefIdx(subPartIdx);2177
+ }2178
+ }2179
2180
- /* determine which motion references the parent CU should search */2181
- SplitData splitCUData;2182
- splitCUData.initSplitCUData();2183
- if (m_param->limitReferences & X265_REF_LIMIT_DEPTH)2184
- {2185
- if (md.bestMode == &md.pred[PRED_SPLIT])2186
- splitCUData.splitRefs = allSplitRefs;2187
- else2188
+ if (m_param->limitModes)2189
{2190
- /* use best merge/inter mode, in case of intra use 2Nx2N inter references */2191
- CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;2192
- uint32_t numPU = cu.getNumPartInter(0);2193
- for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))2194
- splitCUData.splitRefs |= cu.getBestRefIdx(subPartIdx);2195
+ splitCUData.mvCost[0] = md.pred[PRED_2Nx2N].bestME[0][0].mvCost; // L02196
+ splitCUData.mvCost[1] = md.pred[PRED_2Nx2N].bestME[0][1].mvCost; // L12197
+ splitCUData.sa8dCost = md.pred[PRED_2Nx2N].rdCost;2198
}2199
- }2200
2201
- if (m_param->limitModes)2202
- {2203
- splitCUData.mvCost[0] = md.pred[PRED_2Nx2N].bestME[0][0].mvCost; // L02204
- splitCUData.mvCost[1] = md.pred[PRED_2Nx2N].bestME[0][1].mvCost; // L12205
- splitCUData.sa8dCost = md.pred[PRED_2Nx2N].rdCost;2206
+ /* Copy best data to encData CTU and recon */2207
+ md.bestMode->cu.copyToPic(depth);2208
+ md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, parentCTU.m_cuAddr, cuGeom.absPartIdx);2209
}2210
+ else2211
+ {2212
+ if (m_param->bMVType && cuGeom.numPartitions <= 16)2213
+ {2214
+ qprdRefine(parentCTU, cuGeom, qp, qp);2215
2216
- /* Copy best data to encData CTU and recon */2217
- md.bestMode->cu.copyToPic(depth);2218
- md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, parentCTU.m_cuAddr, cuGeom.absPartIdx);2219
+ SplitData splitData[4];2220
+ splitData[0].initSplitCUData();2221
+ splitData[1].initSplitCUData();2222
+ splitData[2].initSplitCUData();2223
+ splitData[3].initSplitCUData();2224
+2225
+ uint32_t allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;2226
+2227
+ splitCUData.initSplitCUData();2228
+ if (m_param->limitReferences & X265_REF_LIMIT_DEPTH)2229
+ {2230
+ if (md.bestMode == &md.pred[PRED_SPLIT])2231
+ splitCUData.splitRefs = allSplitRefs;2232
+ else2233
+ {2234
+ /* use best merge/inter mode, in case of intra use 2Nx2N inter references */2235
+ CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;2236
+ uint32_t numPU = cu.getNumPartInter(0);2237
+ for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))2238
+ splitCUData.splitRefs |= cu.getBestRefIdx(subPartIdx);2239
+ }2240
+ }2241
+2242
+ if (m_param->limitModes)2243
+ {2244
+ splitCUData.mvCost[0] = md.pred[PRED_2Nx2N].bestME[0][0].mvCost; // L02245
+ splitCUData.mvCost[1] = md.pred[PRED_2Nx2N].bestME[0][1].mvCost; // L12246
+ splitCUData.sa8dCost = md.pred[PRED_2Nx2N].rdCost;2247
+ }2248
+ }2249
+ }2250
2251
return splitCUData;2252
}2253
2254
bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);2255
bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;2256
2257
- int split = (m_param->interRefine && cuGeom.log2CUSize == (uint32_t)(g_log2Size[m_param->minCUSize] + 1)2258
- && bDecidedDepth && parentCTU.m_predMode[cuGeom.absPartIdx] == MODE_SKIP);2259
+ int split = (m_param->interRefine && cuGeom.log2CUSize == (uint32_t)(g_log2Size[m_param->minCUSize] + 1) && bDecidedDepth);2260
2261
if (bDecidedDepth)2262
{2263
2264
md.bestMode = &mode;2265
mode.cu.initSubCU(parentCTU, cuGeom, qp);2266
PartSize size = (PartSize)parentCTU.m_partSize[cuGeom.absPartIdx];2267
- if (parentCTU.isIntra(cuGeom.absPartIdx))2268
+ if (parentCTU.isIntra(cuGeom.absPartIdx) && m_param->interRefine < 2)2269
{2270
- if (m_param->intraRefine != 2 || parentCTU.m_lumaIntraDir[cuGeom.absPartIdx] <= 1)2271
+ bool reuseModes = !((m_param->intraRefine == 3) ||2272
+ (m_param->intraRefine == 2 && parentCTU.m_lumaIntraDir[cuGeom.absPartIdx] > DC_IDX));2273
+ if (reuseModes)2274
{2275
memcpy(mode.cu.m_lumaIntraDir, parentCTU.m_lumaIntraDir + cuGeom.absPartIdx, cuGeom.numPartitions);2276
memcpy(mode.cu.m_chromaIntraDir, parentCTU.m_chromaIntraDir + cuGeom.absPartIdx, cuGeom.numPartitions);2277
}2278
checkIntra(mode, cuGeom, size);2279
}2280
- else2281
+ else if (!parentCTU.isIntra(cuGeom.absPartIdx) && m_param->interRefine < 2)2282
{2283
mode.cu.copyFromPic(parentCTU, cuGeom, m_csp, false);2284
uint32_t numPU = parentCTU.getNumPartInter(cuGeom.absPartIdx);2285
for (uint32_t part = 0; part < numPU; part++)2286
{2287
PredictionUnit pu(mode.cu, cuGeom, part);2288
- if (m_param->analysisReuseLevel == 10)2289
+ if (m_param->analysisReuseLevel >= 7)2290
{2291
analysis_inter_data* interDataCTU = (analysis_inter_data*)m_frame->m_analysisData.interData;2292
int cuIdx = (mode.cu.m_cuAddr * parentCTU.m_numPartitions) + cuGeom.absPartIdx;2293
2294
checkDQP(mode, cuGeom);2295
}2296
2297
- if (m_bTryLossless)2298
- tryLossless(cuGeom);2299
+ if (m_param->interRefine < 2)2300
+ {2301
+ if (m_bTryLossless)2302
+ tryLossless(cuGeom);2303
2304
- if (mightSplit)2305
- addSplitFlagCost(*md.bestMode, cuGeom.depth);2306
+ if (mightSplit)2307
+ addSplitFlagCost(*md.bestMode, cuGeom.depth);2308
2309
- if (mightSplit && m_param->rdLevel < 5)2310
- checkDQPForSplitPred(*md.bestMode, cuGeom);2311
+ if (mightSplit && m_param->rdLevel < 5)2312
+ checkDQPForSplitPred(*md.bestMode, cuGeom);2313
+ }2314
+2315
+ if (m_param->bMVType && m_param->analysisReuseLevel == 7)2316
+ {2317
+ for (int list = 0; list < m_slice->isInterB() + 1; list++)2318
+ {2319
+ m_modeFlag[list] = true;2320
+ if (parentCTU.m_skipFlag[list][cuGeom.absPartIdx] == 1 && cuGeom.numPartitions <= 16)2321
+ m_checkMergeAndSkipOnly[list] = true;2322
+ }2323
+ m_param->rdLevel > 4 ? compressInterCU_rd5_6(parentCTU, cuGeom, qp) : compressInterCU_rd0_4(parentCTU, cuGeom, qp);2324
+ for (int list = 0; list < m_slice->isInterB() + 1; list++)2325
+ {2326
+ m_modeFlag[list] = false;2327
+ m_checkMergeAndSkipOnly[list] = false;2328
+ }2329
+ }2330
2331
- if (m_param->interRefine && parentCTU.m_predMode[cuGeom.absPartIdx] == MODE_SKIP && !mode.cu.isSkipped(0))2332
+ if (m_param->interRefine > 1 || (m_param->interRefine && parentCTU.m_predMode[cuGeom.absPartIdx] == MODE_SKIP && !mode.cu.isSkipped(0)))2333
{2334
m_evaluateInter = 1;2335
m_param->rdLevel > 4 ? compressInterCU_rd5_6(parentCTU, cuGeom, qp) : compressInterCU_rd0_4(parentCTU, cuGeom, qp);2336
+ m_evaluateInter = 0;2337
}2338
}2339
if (!bDecidedDepth || split)2340
2341
if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)2342
nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));2343
2344
- int lamdaQP = m_param->analysisReuseLevel == 10 ? nextQP : lqp;2345
+ int lamdaQP = (m_param->analysisReuseLevel >= 7) ? nextQP : lqp;2346
2347
if (split)2348
m_param->rdLevel > 4 ? compressInterCU_rd5_6(parentCTU, childGeom, nextQP) : compressInterCU_rd0_4(parentCTU, childGeom, nextQP);2349
x265_2.5.tar.gz/source/encoder/analysis.h -> x265_2.6.tar.gz/source/encoder/analysis.h
Changed
20
1
2
bool m_bChromaSa8d;3
bool m_bHD;4
5
+ bool m_modeFlag[2];6
+ bool m_checkMergeAndSkipOnly[2];7
+8
Analysis();9
10
bool create(ThreadLocalData* tld);11
12
void qprdRefine(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp, int32_t lqp);13
14
/* full analysis for an I-slice CU */15
- void compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp);16
+ uint64_t compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp);17
18
/* full analysis for a P or B slice CU */19
uint32_t compressInterCU_dist(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp);20
x265_2.5.tar.gz/source/encoder/api.cpp -> x265_2.6.tar.gz/source/encoder/api.cpp
Changed
600
1
2
#include "level.h"3
#include "nal.h"4
#include "bitcost.h"5
-#include "x265-extras.h"6
7
/* multilib namespace reflectors */8
#if LINKED_8BIT9
10
namespace X265_NS {11
#endif12
13
+static const char* summaryCSVHeader =14
+ "Command, Date/Time, Elapsed Time, FPS, Bitrate, "15
+ "Y PSNR, U PSNR, V PSNR, Global PSNR, SSIM, SSIM (dB), "16
+ "I count, I ave-QP, I kbps, I-PSNR Y, I-PSNR U, I-PSNR V, I-SSIM (dB), "17
+ "P count, P ave-QP, P kbps, P-PSNR Y, P-PSNR U, P-PSNR V, P-SSIM (dB), "18
+ "B count, B ave-QP, B kbps, B-PSNR Y, B-PSNR U, B-PSNR V, B-SSIM (dB), "19
+ "MaxCLL, MaxFALL, Version\n";20
+21
x265_encoder *x265_encoder_open(x265_param *p)22
{23
if (!p)24
25
/* Try to open CSV file handle */26
if (encoder->m_param->csvfn)27
{28
- encoder->m_param->csvfpt = x265_csvlog_open(*encoder->m_param, encoder->m_param->csvfn, encoder->m_param->csvLogLevel);29
+ encoder->m_param->csvfpt = x265_csvlog_open(encoder->m_param);30
if (!encoder->m_param->csvfpt)31
{32
x265_log(encoder->m_param, X265_LOG_ERROR, "Unable to open CSV log file <%s>, aborting\n", encoder->m_param->csvfn);33
34
35
x265_param save;36
Encoder* encoder = static_cast<Encoder*>(enc);37
- if (encoder->m_reconfigure || encoder->m_reconfigureRc) /* Reconfigure in progress */38
+ if (encoder->m_latestParam->forceFlush != param_in->forceFlush)39
+ return encoder->reconfigureParam(encoder->m_latestParam, param_in);40
+ bool isReconfigureRc = encoder->isReconfigureRc(encoder->m_latestParam, param_in);41
+ if ((encoder->m_reconfigure && !isReconfigureRc) || (encoder->m_reconfigureRc && isReconfigureRc)) /* Reconfigure in progress */42
return 1;43
memcpy(&save, encoder->m_latestParam, sizeof(x265_param));44
int ret = encoder->reconfigureParam(encoder->m_latestParam, param_in);45
46
if (encoder->m_param->bRepeatHeaders)47
{48
if (encoder->m_scalingList.parseScalingList(encoder->m_latestParam->scalingLists))49
+ {50
+ memcpy(encoder->m_latestParam, &save, sizeof(x265_param));51
return -1;52
+ }53
encoder->m_scalingList.setupQuantMatrices(encoder->m_param->internalCsp);54
}55
else56
{57
x265_log(encoder->m_param, X265_LOG_ERROR, "Repeat headers is turned OFF, cannot reconfigure scalinglists\n");58
+ memcpy(encoder->m_latestParam, &save, sizeof(x265_param));59
return -1;60
}61
}62
- if (encoder->m_reconfigureRc)63
+ if (!isReconfigureRc)64
+ encoder->m_reconfigure = true;65
+ else if (encoder->m_reconfigureRc)66
{67
VPS saveVPS;68
memcpy(&saveVPS.ptl, &encoder->m_vps.ptl, sizeof(saveVPS.ptl));69
70
x265_log(encoder->m_param, X265_LOG_WARNING, "Profile/Level/Tier has changed from %d/%d/%s to %d/%d/%s.Cannot reconfigure rate-control.\n",71
saveVPS.ptl.profileIdc, saveVPS.ptl.levelIdc, saveVPS.ptl.tierFlag ? "High" : "Main", encoder->m_vps.ptl.profileIdc,72
encoder->m_vps.ptl.levelIdc, encoder->m_vps.ptl.tierFlag ? "High" : "Main");73
+ memcpy(encoder->m_latestParam, &save, sizeof(x265_param));74
+ memcpy(&encoder->m_vps.ptl, &saveVPS.ptl, sizeof(saveVPS.ptl));75
encoder->m_reconfigureRc = false;76
}77
}78
- else79
- encoder->m_reconfigure = true;80
encoder->printReconfigureParams();81
}82
return ret;83
84
{85
numEncoded = encoder->encode(pic_in, pic_out);86
}87
- while (numEncoded == 0 && !pic_in && encoder->m_numDelayedPic);88
+ while ((numEncoded == 0 && !pic_in && encoder->m_numDelayedPic && !encoder->m_latestParam->forceFlush) && !encoder->m_externalFlush);89
+ if (numEncoded)90
+ encoder->m_externalFlush = false;91
92
// do not allow reuse of these buffers for more than one picture. The93
// encoder now owns these analysisData buffers.94
95
*pi_nal = 0;96
97
if (numEncoded && encoder->m_param->csvLogLevel)98
- x265_csvlog_frame(encoder->m_param->csvfpt, *encoder->m_param, *pic_out, encoder->m_param->csvLogLevel);99
+ x265_csvlog_frame(encoder->m_param, pic_out);100
101
if (numEncoded < 0)102
encoder->m_aborted = true;103
104
{105
Encoder *encoder = static_cast<Encoder*>(enc);106
x265_stats stats;107
- int padx = encoder->m_sps.conformanceWindow.rightOffset;108
- int pady = encoder->m_sps.conformanceWindow.bottomOffset;109
encoder->fetchStats(&stats, sizeof(stats));110
- const x265_api * api = x265_api_get(0);111
- x265_csvlog_encode(encoder->m_param->csvfpt, api->version_str, *encoder->m_param, padx, pady, stats, encoder->m_param->csvLogLevel, argc, argv);112
+ x265_csvlog_encode(enc, &stats, argc, argv);113
}114
}115
116
117
return 0;118
}119
120
+int x265_get_slicetype_poc_and_scenecut(x265_encoder *enc, int *slicetype, int *poc, int *sceneCut)121
+{122
+ if (!enc)123
+ return -1;124
+ Encoder *encoder = static_cast<Encoder*>(enc);125
+ if (!encoder->copySlicetypePocAndSceneCut(slicetype, poc, sceneCut))126
+ return 0;127
+ return -1;128
+}129
+130
+int x265_get_ref_frame_list(x265_encoder *enc, x265_picyuv** l0, x265_picyuv** l1, int sliceType, int poc)131
+{132
+ if (!enc)133
+ return -1;134
+135
+ Encoder *encoder = static_cast<Encoder*>(enc);136
+ return encoder->getRefFrameList((PicYuv**)l0, (PicYuv**)l1, sliceType, poc);137
+}138
+139
+int x265_set_analysis_data(x265_encoder *enc, x265_analysis_data *analysis_data, int poc, uint32_t cuBytes)140
+{141
+ if (!enc)142
+ return -1;143
+144
+ Encoder *encoder = static_cast<Encoder*>(enc);145
+ if (!encoder->setAnalysisData(analysis_data, poc, cuBytes))146
+ return 0;147
+148
+ return -1;149
+}150
+151
void x265_cleanup(void)152
{153
BitCost::destroy();154
155
pic->userSEI.payloads = NULL;156
pic->userSEI.numPayloads = 0;157
158
- if (param->analysisReuseMode)159
+ if (param->analysisReuseMode || (param->bMVType == AVC_INFO))160
{161
uint32_t widthInCU = (param->sourceWidth + param->maxCUSize - 1) >> param->maxLog2CUSize;162
uint32_t heightInCU = (param->sourceHeight + param->maxCUSize - 1) >> param->maxLog2CUSize;163
164
sizeof(x265_frame_stats),165
&x265_encoder_intra_refresh,166
&x265_encoder_ctu_info,167
+ &x265_get_slicetype_poc_and_scenecut,168
+ &x265_get_ref_frame_list,169
+ &x265_csvlog_open,170
+ &x265_csvlog_frame,171
+ &x265_csvlog_encode,172
+ &x265_dither_image,173
+ &x265_set_analysis_data174
};175
176
typedef const x265_api* (*api_get_func)(int bitDepth);177
178
return &libapi;179
}180
181
+FILE* x265_csvlog_open(const x265_param* param)182
+{183
+ FILE *csvfp = x265_fopen(param->csvfn, "r");184
+ if (csvfp)185
+ {186
+ /* file already exists, re-open for append */187
+ fclose(csvfp);188
+ return x265_fopen(param->csvfn, "ab");189
+ }190
+ else191
+ {192
+ /* new CSV file, write header */193
+ csvfp = x265_fopen(param->csvfn, "wb");194
+ if (csvfp)195
+ {196
+ if (param->csvLogLevel)197
+ {198
+ fprintf(csvfp, "Encode Order, Type, POC, QP, Bits, Scenecut, ");199
+ if (param->csvLogLevel >= 2)200
+ fprintf(csvfp, "I/P cost ratio, ");201
+ if (param->rc.rateControlMode == X265_RC_CRF)202
+ fprintf(csvfp, "RateFactor, ");203
+ if (param->rc.vbvBufferSize)204
+ fprintf(csvfp, "BufferFill, ");205
+ if (param->bEnablePsnr)206
+ fprintf(csvfp, "Y PSNR, U PSNR, V PSNR, YUV PSNR, ");207
+ if (param->bEnableSsim)208
+ fprintf(csvfp, "SSIM, SSIM(dB), ");209
+ fprintf(csvfp, "Latency, ");210
+ fprintf(csvfp, "List 0, List 1");211
+ uint32_t size = param->maxCUSize;212
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)213
+ {214
+ fprintf(csvfp, ", Intra %dx%d DC, Intra %dx%d Planar, Intra %dx%d Ang", size, size, size, size, size, size);215
+ size /= 2;216
+ }217
+ fprintf(csvfp, ", 4x4");218
+ size = param->maxCUSize;219
+ if (param->bEnableRectInter)220
+ {221
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)222
+ {223
+ fprintf(csvfp, ", Inter %dx%d, Inter %dx%d (Rect)", size, size, size, size);224
+ if (param->bEnableAMP)225
+ fprintf(csvfp, ", Inter %dx%d (Amp)", size, size);226
+ size /= 2;227
+ }228
+ }229
+ else230
+ {231
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)232
+ {233
+ fprintf(csvfp, ", Inter %dx%d", size, size);234
+ size /= 2;235
+ }236
+ }237
+ size = param->maxCUSize;238
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)239
+ {240
+ fprintf(csvfp, ", Skip %dx%d", size, size);241
+ size /= 2;242
+ }243
+ size = param->maxCUSize;244
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)245
+ {246
+ fprintf(csvfp, ", Merge %dx%d", size, size);247
+ size /= 2;248
+ }249
+250
+ if (param->csvLogLevel >= 2)251
+ {252
+ fprintf(csvfp, ", Avg Luma Distortion, Avg Chroma Distortion, Avg psyEnergy, Avg Residual Energy,"253
+ " Min Luma Level, Max Luma Level, Avg Luma Level");254
+255
+ if (param->internalCsp != X265_CSP_I400)256
+ fprintf(csvfp, ", Min Cb Level, Max Cb Level, Avg Cb Level, Min Cr Level, Max Cr Level, Avg Cr Level");257
+258
+ /* PU statistics */259
+ size = param->maxCUSize;260
+ for (uint32_t i = 0; i< param->maxLog2CUSize - (uint32_t)g_log2Size[param->minCUSize] + 1; i++)261
+ {262
+ fprintf(csvfp, ", Intra %dx%d", size, size);263
+ fprintf(csvfp, ", Skip %dx%d", size, size);264
+ fprintf(csvfp, ", AMP %d", size);265
+ fprintf(csvfp, ", Inter %dx%d", size, size);266
+ fprintf(csvfp, ", Merge %dx%d", size, size);267
+ fprintf(csvfp, ", Inter %dx%d", size, size / 2);268
+ fprintf(csvfp, ", Merge %dx%d", size, size / 2);269
+ fprintf(csvfp, ", Inter %dx%d", size / 2, size);270
+ fprintf(csvfp, ", Merge %dx%d", size / 2, size);271
+ size /= 2;272
+ }273
+274
+ if ((uint32_t)g_log2Size[param->minCUSize] == 3)275
+ fprintf(csvfp, ", 4x4");276
+277
+ /* detailed performance statistics */278
+ fprintf(csvfp, ", DecideWait (ms), Row0Wait (ms), Wall time (ms), Ref Wait Wall (ms), Total CTU time (ms),"279
+ "Stall Time (ms), Total frame time (ms), Avg WPP, Row Blocks");280
+ }281
+ fprintf(csvfp, "\n");282
+ }283
+ else284
+ fputs(summaryCSVHeader, csvfp);285
+ }286
+ return csvfp;287
+ }288
+}289
+290
+// per frame CSV logging291
+void x265_csvlog_frame(const x265_param* param, const x265_picture* pic)292
+{293
+ if (!param->csvfpt)294
+ return;295
+296
+ const x265_frame_stats* frameStats = &pic->frameData;297
+ fprintf(param->csvfpt, "%d, %c-SLICE, %4d, %2.2lf, %10d, %d,", frameStats->encoderOrder, frameStats->sliceType, frameStats->poc,298
+ frameStats->qp, (int)frameStats->bits, frameStats->bScenecut);299
+ if (param->csvLogLevel >= 2)300
+ fprintf(param->csvfpt, "%.2f,", frameStats->ipCostRatio);301
+ if (param->rc.rateControlMode == X265_RC_CRF)302
+ fprintf(param->csvfpt, "%.3lf,", frameStats->rateFactor);303
+ if (param->rc.vbvBufferSize)304
+ fprintf(param->csvfpt, "%.3lf,", frameStats->bufferFill);305
+ if (param->bEnablePsnr)306
+ fprintf(param->csvfpt, "%.3lf, %.3lf, %.3lf, %.3lf,", frameStats->psnrY, frameStats->psnrU, frameStats->psnrV, frameStats->psnr);307
+ if (param->bEnableSsim)308
+ fprintf(param->csvfpt, " %.6f, %6.3f,", frameStats->ssim, x265_ssim2dB(frameStats->ssim));309
+ fprintf(param->csvfpt, "%d, ", frameStats->frameLatency);310
+ if (frameStats->sliceType == 'I' || frameStats->sliceType == 'i')311
+ fputs(" -, -,", param->csvfpt);312
+ else313
+ {314
+ int i = 0;315
+ while (frameStats->list0POC[i] != -1)316
+ fprintf(param->csvfpt, "%d ", frameStats->list0POC[i++]);317
+ fprintf(param->csvfpt, ",");318
+ if (frameStats->sliceType != 'P')319
+ {320
+ i = 0;321
+ while (frameStats->list1POC[i] != -1)322
+ fprintf(param->csvfpt, "%d ", frameStats->list1POC[i++]);323
+ fprintf(param->csvfpt, ",");324
+ }325
+ else326
+ fputs(" -,", param->csvfpt);327
+ }328
+329
+ if (param->csvLogLevel)330
+ {331
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)332
+ fprintf(param->csvfpt, "%5.2lf%%, %5.2lf%%, %5.2lf%%,", frameStats->cuStats.percentIntraDistribution[depth][0],333
+ frameStats->cuStats.percentIntraDistribution[depth][1],334
+ frameStats->cuStats.percentIntraDistribution[depth][2]);335
+ fprintf(param->csvfpt, "%5.2lf%%", frameStats->cuStats.percentIntraNxN);336
+ if (param->bEnableRectInter)337
+ {338
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)339
+ {340
+ fprintf(param->csvfpt, ", %5.2lf%%, %5.2lf%%", frameStats->cuStats.percentInterDistribution[depth][0],341
+ frameStats->cuStats.percentInterDistribution[depth][1]);342
+ if (param->bEnableAMP)343
+ fprintf(param->csvfpt, ", %5.2lf%%", frameStats->cuStats.percentInterDistribution[depth][2]);344
+ }345
+ }346
+ else347
+ {348
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)349
+ fprintf(param->csvfpt, ", %5.2lf%%", frameStats->cuStats.percentInterDistribution[depth][0]);350
+ }351
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)352
+ fprintf(param->csvfpt, ", %5.2lf%%", frameStats->cuStats.percentSkipCu[depth]);353
+ for (uint32_t depth = 0; depth <= param->maxCUDepth; depth++)354
+ fprintf(param->csvfpt, ", %5.2lf%%", frameStats->cuStats.percentMergeCu[depth]);355
+ }356
+357
+ if (param->csvLogLevel >= 2)358
+ {359
+ fprintf(param->csvfpt, ", %.2lf, %.2lf, %.2lf, %.2lf ", frameStats->avgLumaDistortion,360
+ frameStats->avgChromaDistortion,361
+ frameStats->avgPsyEnergy,362
+ frameStats->avgResEnergy);363
+364
+ fprintf(param->csvfpt, ", %d, %d, %.2lf", frameStats->minLumaLevel, frameStats->maxLumaLevel, frameStats->avgLumaLevel);365
+366
+ if (param->internalCsp != X265_CSP_I400)367
+ {368
+ fprintf(param->csvfpt, ", %d, %d, %.2lf", frameStats->minChromaULevel, frameStats->maxChromaULevel, frameStats->avgChromaULevel);369
+ fprintf(param->csvfpt, ", %d, %d, %.2lf", frameStats->minChromaVLevel, frameStats->maxChromaVLevel, frameStats->avgChromaVLevel);370
+ }371
+372
+ for (uint32_t i = 0; i < param->maxLog2CUSize - (uint32_t)g_log2Size[param->minCUSize] + 1; i++)373
+ {374
+ fprintf(param->csvfpt, ", %.2lf%%", frameStats->puStats.percentIntraPu[i]);375
+ fprintf(param->csvfpt, ", %.2lf%%", frameStats->puStats.percentSkipPu[i]);376
+ fprintf(param->csvfpt, ",%.2lf%%", frameStats->puStats.percentAmpPu[i]);377
+ for (uint32_t j = 0; j < 3; j++)378
+ {379
+ fprintf(param->csvfpt, ", %.2lf%%", frameStats->puStats.percentInterPu[i][j]);380
+ fprintf(param->csvfpt, ", %.2lf%%", frameStats->puStats.percentMergePu[i][j]);381
+ }382
+ }383
+ if ((uint32_t)g_log2Size[param->minCUSize] == 3)384
+ fprintf(param->csvfpt, ",%.2lf%%", frameStats->puStats.percentNxN);385
+386
+ fprintf(param->csvfpt, ", %.1lf, %.1lf, %.1lf, %.1lf, %.1lf, %.1lf, %.1lf,", frameStats->decideWaitTime, frameStats->row0WaitTime,387
+ frameStats->wallTime, frameStats->refWaitWallTime,388
+ frameStats->totalCTUTime, frameStats->stallTime,389
+ frameStats->totalFrameTime);390
+391
+ fprintf(param->csvfpt, " %.3lf, %d", frameStats->avgWPP, frameStats->countRowBlocks);392
+ }393
+ fprintf(param->csvfpt, "\n");394
+ fflush(stderr);395
+}396
+397
+void x265_csvlog_encode(x265_encoder *enc, const x265_stats* stats, int argc, char** argv)398
+{399
+ if (enc)400
+ {401
+ Encoder *encoder = static_cast<Encoder*>(enc);402
+ int padx = encoder->m_sps.conformanceWindow.rightOffset;403
+ int pady = encoder->m_sps.conformanceWindow.bottomOffset;404
+ const x265_api * api = x265_api_get(0);405
+406
+ if (!encoder->m_param->csvfpt)407
+ return;408
+409
+ if (encoder->m_param->csvLogLevel)410
+ {411
+ // adding summary to a per-frame csv log file, so it needs a summary header412
+ fprintf(encoder->m_param->csvfpt, "\nSummary\n");413
+ fputs(summaryCSVHeader, encoder->m_param->csvfpt);414
+ }415
+416
+ // CLI arguments or other417
+ if (argc)418
+ {419
+ fputc('"', encoder->m_param->csvfpt);420
+ for (int i = 1; i < argc; i++)421
+ {422
+ fputc(' ', encoder->m_param->csvfpt);423
+ fputs(argv[i], encoder->m_param->csvfpt);424
+ }425
+ fputc('"', encoder->m_param->csvfpt);426
+ }427
+ else428
+ {429
+ const x265_param* paramTemp = encoder->m_param;430
+ char *opts = x265_param2string((x265_param*)paramTemp, padx, pady);431
+ if (opts)432
+ {433
+ fputc('"', encoder->m_param->csvfpt);434
+ fputs(opts, encoder->m_param->csvfpt);435
+ fputc('"', encoder->m_param->csvfpt);436
+ }437
+ }438
+439
+ // current date and time440
+ time_t now;441
+ struct tm* timeinfo;442
+ time(&now);443
+ timeinfo = localtime(&now);444
+ char buffer[200];445
+ strftime(buffer, 128, "%c", timeinfo);446
+ fprintf(encoder->m_param->csvfpt, ", %s, ", buffer);447
+448
+ // elapsed time, fps, bitrate449
+ fprintf(encoder->m_param->csvfpt, "%.2f, %.2f, %.2f,",450
+ stats->elapsedEncodeTime, stats->encodedPictureCount / stats->elapsedEncodeTime, stats->bitrate);451
+452
+ if (encoder->m_param->bEnablePsnr)453
+ fprintf(encoder->m_param->csvfpt, " %.3lf, %.3lf, %.3lf, %.3lf,",454
+ stats->globalPsnrY / stats->encodedPictureCount, stats->globalPsnrU / stats->encodedPictureCount,455
+ stats->globalPsnrV / stats->encodedPictureCount, stats->globalPsnr);456
+ else457
+ fprintf(encoder->m_param->csvfpt, " -, -, -, -,");458
+ if (encoder->m_param->bEnableSsim)459
+ fprintf(encoder->m_param->csvfpt, " %.6f, %6.3f,", stats->globalSsim, x265_ssim2dB(stats->globalSsim));460
+ else461
+ fprintf(encoder->m_param->csvfpt, " -, -,");462
+463
+ if (stats->statsI.numPics)464
+ {465
+ fprintf(encoder->m_param->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsI.numPics, stats->statsI.avgQp, stats->statsI.bitrate);466
+ if (encoder->m_param->bEnablePsnr)467
+ fprintf(encoder->m_param->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsI.psnrY, stats->statsI.psnrU, stats->statsI.psnrV);468
+ else469
+ fprintf(encoder->m_param->csvfpt, " -, -, -,");470
+ if (encoder->m_param->bEnableSsim)471
+ fprintf(encoder->m_param->csvfpt, " %.3lf,", stats->statsI.ssim);472
+ else473
+ fprintf(encoder->m_param->csvfpt, " -,");474
+ }475
+ else476
+ fprintf(encoder->m_param->csvfpt, " -, -, -, -, -, -, -,");477
+478
+ if (stats->statsP.numPics)479
+ {480
+ fprintf(encoder->m_param->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsP.numPics, stats->statsP.avgQp, stats->statsP.bitrate);481
+ if (encoder->m_param->bEnablePsnr)482
+ fprintf(encoder->m_param->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsP.psnrY, stats->statsP.psnrU, stats->statsP.psnrV);483
+ else484
+ fprintf(encoder->m_param->csvfpt, " -, -, -,");485
+ if (encoder->m_param->bEnableSsim)486
+ fprintf(encoder->m_param->csvfpt, " %.3lf,", stats->statsP.ssim);487
+ else488
+ fprintf(encoder->m_param->csvfpt, " -,");489
+ }490
+ else491
+ fprintf(encoder->m_param->csvfpt, " -, -, -, -, -, -, -,");492
+493
+ if (stats->statsB.numPics)494
+ {495
+ fprintf(encoder->m_param->csvfpt, " %-6u, %2.2lf, %-8.2lf,", stats->statsB.numPics, stats->statsB.avgQp, stats->statsB.bitrate);496
+ if (encoder->m_param->bEnablePsnr)497
+ fprintf(encoder->m_param->csvfpt, " %.3lf, %.3lf, %.3lf,", stats->statsB.psnrY, stats->statsB.psnrU, stats->statsB.psnrV);498
+ else499
+ fprintf(encoder->m_param->csvfpt, " -, -, -,");500
+ if (encoder->m_param->bEnableSsim)501
+ fprintf(encoder->m_param->csvfpt, " %.3lf,", stats->statsB.ssim);502
+ else503
+ fprintf(encoder->m_param->csvfpt, " -,");504
+ }505
+ else506
+ fprintf(encoder->m_param->csvfpt, " -, -, -, -, -, -, -,");507
+508
+ fprintf(encoder->m_param->csvfpt, " %-6u, %-6u, %s\n", stats->maxCLL, stats->maxFALL, api->version_str);509
+ }510
+}511
+512
+/* The dithering algorithm is based on Sierra-2-4A error diffusion.513
+ * We convert planes in place (without allocating a new buffer). */514
+static void ditherPlane(uint16_t *src, int srcStride, int width, int height, int16_t *errors, int bitDepth)515
+{516
+ const int lShift = 16 - bitDepth;517
+ const int rShift = 16 - bitDepth + 2;518
+ const int half = (1 << (16 - bitDepth + 1));519
+ const int pixelMax = (1 << bitDepth) - 1;520
+521
+ memset(errors, 0, (width + 1) * sizeof(int16_t));522
+523
+ if (bitDepth == 8)524
+ {525
+ for (int y = 0; y < height; y++, src += srcStride)526
+ {527
+ uint8_t* dst = (uint8_t *)src;528
+ int16_t err = 0;529
+ for (int x = 0; x < width; x++)530
+ {531
+ err = err * 2 + errors[x] + errors[x + 1];532
+ int tmpDst = x265_clip3(0, pixelMax, ((src[x] << 2) + err + half) >> rShift);533
+ errors[x] = err = (int16_t)(src[x] - (tmpDst << lShift));534
+ dst[x] = (uint8_t)tmpDst;535
+ }536
+ }537
+ }538
+ else539
+ {540
+ for (int y = 0; y < height; y++, src += srcStride)541
+ {542
+ int16_t err = 0;543
+ for (int x = 0; x < width; x++)544
+ {545
+ err = err * 2 + errors[x] + errors[x + 1];546
+ int tmpDst = x265_clip3(0, pixelMax, ((src[x] << 2) + err + half) >> rShift);547
+ errors[x] = err = (int16_t)(src[x] - (tmpDst << lShift));548
+ src[x] = (uint16_t)tmpDst;549
+ }550
+ }551
+ }552
+}553
+554
+void x265_dither_image(x265_picture* picIn, int picWidth, int picHeight, int16_t *errorBuf, int bitDepth)555
+{556
+ const x265_api* api = x265_api_get(0);557
+558
+ if (sizeof(x265_picture) != api->sizeof_picture)559
+ {560
+ fprintf(stderr, "extras [error]: structure size skew, unable to dither\n");561
+ return;562
+ }563
+564
+ if (picIn->bitDepth <= 8)565
+ {566
+ fprintf(stderr, "extras [error]: dither support enabled only for input bitdepth > 8\n");567
+ return;568
+ }569
+570
+ if (picIn->bitDepth == bitDepth)571
+ {572
+ fprintf(stderr, "extras[error]: dither support enabled only if encoder depth is different from picture depth\n");573
+ return;574
+ }575
+576
+ /* This portion of code is from readFrame in x264. */577
+ for (int i = 0; i < x265_cli_csps[picIn->colorSpace].planes; i++)578
+ {579
+ if (picIn->bitDepth < 16)580
+ {581
+ /* upconvert non 16bit high depth planes to 16bit */582
+ uint16_t *plane = (uint16_t*)picIn->planes[i];583
+ uint32_t pixelCount = x265_picturePlaneSize(picIn->colorSpace, picWidth, picHeight, i);584
+ int lShift = 16 - picIn->bitDepth;585
+586
+ /* This loop assumes width is equal to stride which587
+ * happens to be true for file reader outputs */588
+ for (uint32_t j = 0; j < pixelCount; j++)589
+ plane[j] = plane[j] << lShift;590
+ }591
+592
+ int height = (int)(picHeight >> x265_cli_csps[picIn->colorSpace].height[i]);593
+ int width = (int)(picWidth >> x265_cli_csps[picIn->colorSpace].width[i]);594
+595
+ ditherPlane(((uint16_t*)picIn->planes[i]), picIn->stride[i] / 2, width, height, errorBuf, bitDepth);596
+ }597
+}598
+599
} /* end namespace or extern "C" */600
x265_2.5.tar.gz/source/encoder/encoder.cpp -> x265_2.6.tar.gz/source/encoder/encoder.cpp
Changed
748
1
2
const char g_sliceTypeToChar[] = {'B', 'P', 'I'};3
}4
5
+/* Threshold for motion vection, based on expermental result.6
+ * TODO: come up an algorithm for adoptive threshold */7
+8
+#define MVTHRESHOLD 109
+#define PU_2Nx2N 110
+11
static const char* defaultAnalysisFileName = "x265_analysis.dat";12
13
using namespace X265_NS;14
15
16
#if ENABLE_HDR10_PLUS17
m_hdr10plus_api = hdr10plus_api_get();18
- numCimInfo = 0;19
- cim = NULL;20
+ m_numCimInfo = 0;21
+ m_cim = NULL;22
#endif23
24
m_prevTonemapPayload.payload = NULL;25
26
}27
}28
}29
-30
- m_bZeroLatency = !m_param->bframes && !m_param->lookaheadDepth && m_param->frameNumThreads == 1;31
-32
+ m_bZeroLatency = !m_param->bframes && !m_param->lookaheadDepth && m_param->frameNumThreads == 1 && m_param->maxSlices == 1;33
m_aborted |= parseLambdaFile(m_param);34
35
m_encodeStartTime = x265_mdate();36
37
m_nalList.m_annexB = !!m_param->bAnnexB;38
39
m_emitCLLSEI = p->maxCLL || p->maxFALL;40
+41
+#if ENABLE_HDR10_PLUS42
+ if (m_bToneMap)43
+ m_numCimInfo = m_hdr10plus_api->hdr10plus_json_to_movie_cim(m_param->toneMapFile, m_cim);44
+#endif45
}46
47
void Encoder::stopJobs()48
49
}50
}51
52
+int Encoder::copySlicetypePocAndSceneCut(int *slicetype, int *poc, int *sceneCut)53
+{54
+ Frame *FramePtr = m_dpb->m_picList.getCurFrame();55
+ if (FramePtr != NULL)56
+ {57
+ *slicetype = FramePtr->m_lowres.sliceType;58
+ *poc = FramePtr->m_encData->m_slice->m_poc;59
+ *sceneCut = FramePtr->m_lowres.bScenecut;60
+ }61
+ else62
+ {63
+ x265_log(NULL, X265_LOG_WARNING, "Frame is still in lookahead pipeline, this API must be called after (poc >= lookaheadDepth + bframes + 2) condition check\n");64
+ return -1;65
+ }66
+ return 0;67
+}68
+69
+int Encoder::getRefFrameList(PicYuv** l0, PicYuv** l1, int sliceType, int poc)70
+{71
+ if (!(IS_X265_TYPE_I(sliceType)))72
+ {73
+ Frame *framePtr = m_dpb->m_picList.getPOC(poc);74
+ if (framePtr != NULL)75
+ {76
+ for (int j = 0; j < framePtr->m_encData->m_slice->m_numRefIdx[0]; j++) // check only for --ref=n number of frames.77
+ {78
+ if (framePtr->m_encData->m_slice->m_refFrameList[0][j] && framePtr->m_encData->m_slice->m_refFrameList[0][j]->m_reconPic != NULL)79
+ {80
+ int l0POC = framePtr->m_encData->m_slice->m_refFrameList[0][j]->m_poc;81
+ Frame* l0Fp = m_dpb->m_picList.getPOC(l0POC);82
+ if (l0Fp->m_reconPic->m_picOrg[0] == NULL)83
+ l0Fp->m_reconEncoded.wait(); /* If recon is not ready, current frame encoder need to wait. */84
+ l0[j] = l0Fp->m_reconPic;85
+ }86
+ }87
+ for (int j = 0; j < framePtr->m_encData->m_slice->m_numRefIdx[1]; j++) // check only for --ref=n number of frames.88
+ {89
+ if (framePtr->m_encData->m_slice->m_refFrameList[1][j] && framePtr->m_encData->m_slice->m_refFrameList[1][j]->m_reconPic != NULL)90
+ {91
+ int l1POC = framePtr->m_encData->m_slice->m_refFrameList[1][j]->m_poc;92
+ Frame* l1Fp = m_dpb->m_picList.getPOC(l1POC);93
+ if (l1Fp->m_reconPic->m_picOrg[0] == NULL)94
+ l1Fp->m_reconEncoded.wait(); /* If recon is not ready, current frame encoder need to wait. */95
+ l1[j] = l1Fp->m_reconPic;96
+ }97
+ }98
+ }99
+ else100
+ x265_log(NULL, X265_LOG_WARNING, "Refrence List is not in piclist\n");101
+ }102
+ else103
+ {104
+ x265_log(NULL, X265_LOG_ERROR, "I frames does not have a refrence List\n");105
+ return -1;106
+ }107
+ return 0;108
+}109
+110
+int Encoder::setAnalysisDataAfterZScan(x265_analysis_data *analysis_data, Frame* curFrame)111
+{112
+ int mbImageWidth, mbImageHeight;113
+ mbImageWidth = (curFrame->m_fencPic->m_picWidth + 16 - 1) >> 4; //AVC block sizes114
+ mbImageHeight = (curFrame->m_fencPic->m_picHeight + 16 - 1) >> 4;115
+ if (analysis_data->sliceType == X265_TYPE_IDR || analysis_data->sliceType == X265_TYPE_I)116
+ {117
+ curFrame->m_analysisData.sliceType = X265_TYPE_I;118
+ if (m_param->analysisReuseLevel < 7)119
+ return -1;120
+ curFrame->m_analysisData.numPartitions = m_param->num4x4Partitions;121
+ int num16x16inCUWidth = m_param->maxCUSize >> 4;122
+ uint32_t ctuAddr, offset, cuPos;123
+ analysis_intra_data * intraData = (analysis_intra_data *)curFrame->m_analysisData.intraData;124
+ analysis_intra_data * srcIntraData = (analysis_intra_data *)analysis_data->intraData;125
+ for (int i = 0; i < mbImageHeight; i++)126
+ {127
+ for (int j = 0; j < mbImageWidth; j++)128
+ {129
+ int mbIndex = j + i * mbImageWidth;130
+ ctuAddr = (j / num16x16inCUWidth + ((i / num16x16inCUWidth) * (mbImageWidth / num16x16inCUWidth)));131
+ offset = ((i % num16x16inCUWidth) << 5) + ((j % num16x16inCUWidth) << 4);132
+ if ((j % 4 >= 2) && m_param->maxCUSize == 64)133
+ offset += (2 * 16);134
+ if ((i % 4 >= 2) && m_param->maxCUSize == 64)135
+ offset += (2 * 32);136
+ cuPos = ctuAddr * curFrame->m_analysisData.numPartitions + offset;137
+ memcpy(&(intraData)->depth[cuPos], &(srcIntraData)->depth[mbIndex * 16], 16);138
+ memcpy(&(intraData)->chromaModes[cuPos], &(srcIntraData)->chromaModes[mbIndex * 16], 16);139
+ memcpy(&(intraData)->partSizes[cuPos], &(srcIntraData)->partSizes[mbIndex * 16], 16);140
+ memcpy(&(intraData)->partSizes[cuPos], &(srcIntraData)->partSizes[mbIndex * 16], 16);141
+ }142
+ }143
+ memcpy(&(intraData)->modes, (srcIntraData)->modes, curFrame->m_analysisData.numPartitions * analysis_data->numCUsInFrame);144
+ }145
+ else146
+ {147
+ uint32_t numDir = analysis_data->sliceType == X265_TYPE_P ? 1 : 2;148
+ if (m_param->analysisReuseLevel < 7)149
+ return -1;150
+ curFrame->m_analysisData.numPartitions = m_param->num4x4Partitions;151
+ int num16x16inCUWidth = m_param->maxCUSize >> 4;152
+ uint32_t ctuAddr, offset, cuPos;153
+ analysis_inter_data * interData = (analysis_inter_data *)curFrame->m_analysisData.interData;154
+ analysis_inter_data * srcInterData = (analysis_inter_data*)analysis_data->interData;155
+ for (int i = 0; i < mbImageHeight; i++)156
+ {157
+ for (int j = 0; j < mbImageWidth; j++)158
+ {159
+ int mbIndex = j + i * mbImageWidth;160
+ ctuAddr = (j / num16x16inCUWidth + ((i / num16x16inCUWidth) * (mbImageWidth / num16x16inCUWidth)));161
+ offset = ((i % num16x16inCUWidth) << 5) + ((j % num16x16inCUWidth) << 4);162
+ if ((j % 4 >= 2) && m_param->maxCUSize == 64)163
+ offset += (2 * 16);164
+ if ((i % 4 >= 2) && m_param->maxCUSize == 64)165
+ offset += (2 * 32);166
+ cuPos = ctuAddr * curFrame->m_analysisData.numPartitions + offset;167
+ memcpy(&(interData)->depth[cuPos], &(srcInterData)->depth[mbIndex * 16], 16);168
+ memcpy(&(interData)->modes[cuPos], &(srcInterData)->modes[mbIndex * 16], 16);169
+170
+ memcpy(&(interData)->partSize[cuPos], &(srcInterData)->partSize[mbIndex * 16], 16);171
+172
+ int bytes = curFrame->m_analysisData.numPartitions >> ((srcInterData)->depth[mbIndex * 16] * 2);173
+ int cuCount = 1;174
+ if (bytes < 16)175
+ cuCount = 4;176
+ for (int cuI = 0; cuI < cuCount; cuI++)177
+ {178
+ int numPU = nbPartsTable[(srcInterData)->partSize[mbIndex * 16 + cuI * bytes]];179
+ for (int pu = 0; pu < numPU; pu++)180
+ {181
+ int cuOffset = cuI * bytes + pu;182
+ (interData)->mergeFlag[cuPos + cuOffset] = (srcInterData)->mergeFlag[(mbIndex * 16) + cuOffset];183
+184
+ (interData)->interDir[cuPos + cuOffset] = (srcInterData)->interDir[(mbIndex * 16) + cuOffset];185
+ for (uint32_t k = 0; k < numDir; k++)186
+ {187
+ (interData)->mvpIdx[k][cuPos + cuOffset] = (srcInterData)->mvpIdx[k][(mbIndex * 16) + cuOffset];188
+ (interData)->refIdx[k][cuPos + cuOffset] = (srcInterData)->refIdx[k][(mbIndex * 16) + cuOffset];189
+ memcpy(&(interData)->mv[k][cuPos + cuOffset], &(srcInterData)->mv[k][(mbIndex * 16) + cuOffset], sizeof(MV));190
+ if (m_param->analysisReuseLevel == 7)191
+ {192
+ int mv_x = ((analysis_inter_data *)curFrame->m_analysisData.interData)->mv[k][(mbIndex * 16) + cuOffset].x;193
+ int mv_y = ((analysis_inter_data *)curFrame->m_analysisData.interData)->mv[k][(mbIndex * 16) + cuOffset].y;194
+ double mv = sqrt(mv_x*mv_x + mv_y*mv_y);195
+ if (numPU == PU_2Nx2N && ((srcInterData)->depth[cuPos + cuOffset] == (m_param->maxCUSize >> 5)) && mv <= MVTHRESHOLD)196
+ memset(&curFrame->m_analysisData.modeFlag[k][cuPos + cuOffset], 1, bytes);197
+ }198
+ }199
+ }200
+ }201
+ }202
+ }203
+ }204
+ return 0;205
+}206
+207
+int Encoder::setAnalysisData(x265_analysis_data *analysis_data, int poc, uint32_t cuBytes)208
+{209
+ uint32_t widthInCU = (m_param->sourceWidth + m_param->maxCUSize - 1) >> m_param->maxLog2CUSize;210
+ uint32_t heightInCU = (m_param->sourceHeight + m_param->maxCUSize - 1) >> m_param->maxLog2CUSize;211
+212
+ Frame* curFrame = m_dpb->m_picList.getPOC(poc);213
+ if (curFrame != NULL)214
+ {215
+ curFrame->m_analysisData = (*analysis_data);216
+ curFrame->m_analysisData.numCUsInFrame = widthInCU * heightInCU;217
+ curFrame->m_analysisData.numPartitions = m_param->num4x4Partitions;218
+ allocAnalysis(&curFrame->m_analysisData);219
+ if (m_param->maxCUSize == 16)220
+ {221
+ if (analysis_data->sliceType == X265_TYPE_IDR || analysis_data->sliceType == X265_TYPE_I)222
+ {223
+ curFrame->m_analysisData.sliceType = X265_TYPE_I;224
+ if (m_param->analysisReuseLevel < 2)225
+ return -1;226
+227
+ curFrame->m_analysisData.numPartitions = m_param->num4x4Partitions;228
+ size_t count = 0;229
+ analysis_intra_data * currIntraData = (analysis_intra_data *)curFrame->m_analysisData.intraData;230
+ analysis_intra_data * intraData = (analysis_intra_data *)analysis_data->intraData;231
+ for (uint32_t d = 0; d < cuBytes; d++)232
+ {233
+ int bytes = curFrame->m_analysisData.numPartitions >> ((intraData)->depth[d] * 2);234
+ memset(&(currIntraData)->depth[count], (intraData)->depth[d], bytes);235
+ memset(&(currIntraData)->chromaModes[count], (intraData)->chromaModes[d], bytes);236
+ memset(&(currIntraData)->partSizes[count], (intraData)->partSizes[d], bytes);237
+ memset(&(currIntraData)->partSizes[count], (intraData)->partSizes[d], bytes);238
+ count += bytes;239
+ }240
+ memcpy(&(currIntraData)->modes, (intraData)->modes, curFrame->m_analysisData.numPartitions * analysis_data->numCUsInFrame);241
+ }242
+ else243
+ {244
+ uint32_t numDir = analysis_data->sliceType == X265_TYPE_P ? 1 : 2;245
+ if (m_param->analysisReuseLevel < 2)246
+ return -1;247
+248
+ curFrame->m_analysisData.numPartitions = m_param->num4x4Partitions;249
+ size_t count = 0;250
+ analysis_inter_data * currInterData = (analysis_inter_data *)curFrame->m_analysisData.interData;251
+ analysis_inter_data * interData = (analysis_inter_data *)analysis_data->interData;252
+ for (uint32_t d = 0; d < cuBytes; d++)253
+ {254
+ int bytes = curFrame->m_analysisData.numPartitions >> ((interData)->depth[d] * 2);255
+ memset(&(currInterData)->depth[count], (interData)->depth[d], bytes);256
+ memset(&(currInterData)->modes[count], (interData)->modes[d], bytes);257
+ memcpy(&(currInterData)->sadCost[count], &((analysis_inter_data*)analysis_data->interData)->sadCost[d], bytes);258
+ if (m_param->analysisReuseLevel > 4)259
+ {260
+ memset(&(currInterData)->partSize[count], (interData)->partSize[d], bytes);261
+ int numPU = nbPartsTable[(currInterData)->partSize[d]];262
+ for (int pu = 0; pu < numPU; pu++, d++)263
+ {264
+ (currInterData)->mergeFlag[count + pu] = (interData)->mergeFlag[d];265
+ if (m_param->analysisReuseLevel >= 7)266
+ {267
+ (currInterData)->interDir[count + pu] = (interData)->interDir[d];268
+ for (uint32_t i = 0; i < numDir; i++)269
+ {270
+ (currInterData)->mvpIdx[i][count + pu] = (interData)->mvpIdx[i][d];271
+ (currInterData)->refIdx[i][count + pu] = (interData)->refIdx[i][d];272
+ memcpy(&(currInterData)->mv[i][count + pu], &(interData)->mv[i][d], sizeof(MV));273
+ if (m_param->analysisReuseLevel == 7)274
+ {275
+ int mv_x = ((analysis_inter_data *)curFrame->m_analysisData.interData)->mv[i][count + pu].x;276
+ int mv_y = ((analysis_inter_data *)curFrame->m_analysisData.interData)->mv[i][count + pu].y;277
+ double mv = sqrt(mv_x*mv_x + mv_y*mv_y);278
+ if (numPU == PU_2Nx2N && m_param->num4x4Partitions <= 16 && mv <= MVTHRESHOLD)279
+ memset(&curFrame->m_analysisData.modeFlag[i][count + pu], 1, bytes);280
+ }281
+ }282
+ }283
+ }284
+ }285
+ count += bytes;286
+ }287
+ }288
+ }289
+ else290
+ setAnalysisDataAfterZScan(analysis_data, curFrame);291
+292
+ curFrame->m_copyMVType.trigger();293
+ return 0;294
+ }295
+ return -1;296
+}297
+298
void Encoder::destroy()299
{300
#if ENABLE_HDR10_PLUS301
- m_hdr10plus_api->hdr10plus_clear_movie(cim, numCimInfo);302
+ if (m_bToneMap)303
+ m_hdr10plus_api->hdr10plus_clear_movie(m_cim, m_numCimInfo);304
#endif305
306
if (m_exportedPic)307
308
}309
if (pic_in)310
{311
+ if (m_latestParam->forceFlush == 1)312
+ {313
+ m_lookahead->setLookaheadQueue();314
+ m_latestParam->forceFlush = 0;315
+ }316
+ if (m_latestParam->forceFlush == 2)317
+ {318
+ m_lookahead->m_filled = false;319
+ m_latestParam->forceFlush = 0;320
+ }321
+322
x265_sei_payload toneMap;323
toneMap.payload = NULL;324
#if ENABLE_HDR10_PLUS325
if (m_bToneMap)326
{327
- if (pic_in->poc == 0)328
- numCimInfo = m_hdr10plus_api->hdr10plus_json_to_movie_cim(m_param->toneMapFile, cim);329
- if (pic_in->poc < numCimInfo)330
+ int currentPOC = m_pocLast + 1;331
+ if (currentPOC < m_numCimInfo)332
{333
int32_t i = 0;334
toneMap.payloadSize = 0;335
- while (cim[pic_in->poc][i] == 0xFF)336
- toneMap.payloadSize += cim[pic_in->poc][i++];337
- toneMap.payloadSize += cim[pic_in->poc][i++];338
+ while (m_cim[currentPOC][i] == 0xFF)339
+ toneMap.payloadSize += m_cim[currentPOC][i++];340
+ toneMap.payloadSize += m_cim[currentPOC][i];341
342
toneMap.payload = (uint8_t*)x265_malloc(sizeof(uint8_t) * toneMap.payloadSize);343
toneMap.payloadType = USER_DATA_REGISTERED_ITU_T_T35;344
- memcpy(toneMap.payload, cim[pic_in->poc] + i, toneMap.payloadSize);345
+ memcpy(toneMap.payload, &m_cim[currentPOC][i+1], toneMap.payloadSize);346
}347
}348
#endif349
350
351
if (inFrame->m_userSEI.numPayloads)352
{353
- inFrame->m_userSEI.payloads = new x265_sei_payload[numPayloads];354
+ if (!inFrame->m_userSEI.payloads)355
+ {356
+ inFrame->m_userSEI.payloads = new x265_sei_payload[numPayloads];357
+ for (int i = 0; i < numPayloads; i++)358
+ inFrame->m_userSEI.payloads[i].payload = NULL;359
+ }360
for (int i = 0; i < numPayloads; i++)361
{362
x265_sei_payload input;363
364
input = pic_in->userSEI.payloads[i];365
int size = inFrame->m_userSEI.payloads[i].payloadSize = input.payloadSize;366
inFrame->m_userSEI.payloads[i].payloadType = input.payloadType;367
- inFrame->m_userSEI.payloads[i].payload = new uint8_t[size];368
+ if (!inFrame->m_userSEI.payloads[i].payload)369
+ inFrame->m_userSEI.payloads[i].payload = new uint8_t[size];370
memcpy(inFrame->m_userSEI.payloads[i].payload, input.payload, size);371
}372
if (toneMap.payload)373
374
{375
/* readAnalysisFile reads analysis data for the frame and allocates memory based on slicetype */376
readAnalysisFile(&inFrame->m_analysisData, inFrame->m_poc, pic_in);377
+ inFrame->m_poc = inFrame->m_analysisData.poc;378
sliceType = inFrame->m_analysisData.sliceType;379
inFrame->m_lowres.bScenecut = !!inFrame->m_analysisData.bScenecut;380
inFrame->m_lowres.satdCost = inFrame->m_analysisData.satdCost;381
+ if (m_param->bDisableLookahead)382
+ {383
+ inFrame->m_lowres.sliceType = sliceType;384
+ inFrame->m_lowres.bKeyframe = !!inFrame->m_analysisData.lookahead.keyframe;385
+ inFrame->m_lowres.bLastMiniGopBFrame = !!inFrame->m_analysisData.lookahead.lastMiniGopBFrame;386
+ int vbvCount = m_param->lookaheadDepth + m_param->bframes + 2;387
+ for (int index = 0; index < vbvCount; index++)388
+ {389
+ inFrame->m_lowres.plannedSatd[index] = inFrame->m_analysisData.lookahead.plannedSatd[index];390
+ inFrame->m_lowres.plannedType[index] = inFrame->m_analysisData.lookahead.plannedType[index];391
+ }392
+ }393
}394
if (m_param->bUseRcStats && pic_in->rcData)395
{396
397
m_lookahead->addPicture(*inFrame, sliceType);398
m_numDelayedPic++;399
}400
+ else if (m_latestParam->forceFlush == 2)401
+ m_lookahead->m_filled = true;402
else403
m_lookahead->flush();404
405
406
x265_frame_stats* frameData = NULL;407
408
/* Free up pic_in->analysisData since it has already been used */409
- if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD)410
+ if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD || (m_param->bMVType && slice->m_sliceType != I_SLICE))411
freeAnalysis(&outFrame->m_analysisData);412
413
if (pic_out)414
415
pic_out->analysisData.poc = pic_out->poc;416
pic_out->analysisData.sliceType = pic_out->sliceType;417
pic_out->analysisData.bScenecut = outFrame->m_lowres.bScenecut;418
- pic_out->analysisData.satdCost = outFrame->m_lowres.satdCost; 419
+ pic_out->analysisData.satdCost = outFrame->m_lowres.satdCost;420
pic_out->analysisData.numCUsInFrame = outFrame->m_analysisData.numCUsInFrame;421
pic_out->analysisData.numPartitions = outFrame->m_analysisData.numPartitions;422
pic_out->analysisData.wt = outFrame->m_analysisData.wt;423
pic_out->analysisData.interData = outFrame->m_analysisData.interData;424
pic_out->analysisData.intraData = outFrame->m_analysisData.intraData;425
+ pic_out->analysisData.modeFlag[0] = outFrame->m_analysisData.modeFlag[0];426
+ pic_out->analysisData.modeFlag[1] = outFrame->m_analysisData.modeFlag[1];427
+ if (m_param->bDisableLookahead)428
+ {429
+ int factor = 1;430
+ if (m_param->scaleFactor)431
+ factor = m_param->scaleFactor * 2;432
+ pic_out->analysisData.numCuInHeight = outFrame->m_analysisData.numCuInHeight;433
+ pic_out->analysisData.lookahead.dts = outFrame->m_dts;434
+ pic_out->analysisData.satdCost *= factor;435
+ pic_out->analysisData.lookahead.keyframe = outFrame->m_lowres.bKeyframe;436
+ pic_out->analysisData.lookahead.lastMiniGopBFrame = outFrame->m_lowres.bLastMiniGopBFrame;437
+ int vbvCount = m_param->lookaheadDepth + m_param->bframes + 2;438
+ for (int index = 0; index < vbvCount; index++)439
+ {440
+ pic_out->analysisData.lookahead.plannedSatd[index] = outFrame->m_lowres.plannedSatd[index] * factor;441
+ pic_out->analysisData.lookahead.plannedType[index] = outFrame->m_lowres.plannedType[index];442
+ }443
+ for (uint32_t index = 0; index < pic_out->analysisData.numCuInHeight; index++)444
+ {445
+ outFrame->m_analysisData.lookahead.intraSatdForVbv[index] = outFrame->m_encData->m_rowStat[index].intraSatdForVbv * factor;446
+ outFrame->m_analysisData.lookahead.satdForVbv[index] = outFrame->m_encData->m_rowStat[index].satdForVbv * factor;447
+ }448
+ pic_out->analysisData.lookahead.intraSatdForVbv = outFrame->m_analysisData.lookahead.intraSatdForVbv;449
+ pic_out->analysisData.lookahead.satdForVbv = outFrame->m_analysisData.lookahead.satdForVbv;450
+ for (uint32_t index = 0; index < pic_out->analysisData.numCUsInFrame; index++)451
+ {452
+ outFrame->m_analysisData.lookahead.intraVbvCost[index] = outFrame->m_encData->m_cuStat[index].intraVbvCost * factor;453
+ outFrame->m_analysisData.lookahead.vbvCost[index] = outFrame->m_encData->m_cuStat[index].vbvCost * factor;454
+ }455
+ pic_out->analysisData.lookahead.intraVbvCost = outFrame->m_analysisData.lookahead.intraVbvCost;456
+ pic_out->analysisData.lookahead.vbvCost = outFrame->m_analysisData.lookahead.vbvCost;457
+ }458
writeAnalysisFile(&pic_out->analysisData, *outFrame->m_encData);459
if (m_param->bUseAnalysisFile)460
freeAnalysis(&pic_out->analysisData);461
462
slice->m_maxNumMergeCand = m_param->maxNumMergeCand;463
slice->m_endCUAddr = slice->realEndAddress(m_sps.numCUsInFrame * m_param->num4x4Partitions);464
}465
-466
+ if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->bDisableLookahead)467
+ {468
+ frameEnc->m_dts = frameEnc->m_analysisData.lookahead.dts;469
+ for (uint32_t index = 0; index < frameEnc->m_analysisData.numCuInHeight; index++)470
+ {471
+ frameEnc->m_encData->m_rowStat[index].intraSatdForVbv = frameEnc->m_analysisData.lookahead.intraSatdForVbv[index];472
+ frameEnc->m_encData->m_rowStat[index].satdForVbv = frameEnc->m_analysisData.lookahead.satdForVbv[index];473
+ }474
+ for (uint32_t index = 0; index < frameEnc->m_analysisData.numCUsInFrame; index++)475
+ {476
+ frameEnc->m_encData->m_cuStat[index].intraVbvCost = frameEnc->m_analysisData.lookahead.intraVbvCost[index];477
+ frameEnc->m_encData->m_cuStat[index].vbvCost = frameEnc->m_analysisData.lookahead.vbvCost[index];478
+ }479
+ }480
if (m_param->searchMethod == X265_SEA && frameEnc->m_lowres.sliceType != X265_TYPE_B)481
{482
int padX = m_param->maxCUSize + 32;483
484
frameEnc->m_encData->m_slice->m_iNumRPSInSPS = m_sps.spsrpsNum;485
486
curEncoder->m_rce.encodeOrder = frameEnc->m_encodeOrder = m_encodedFrameNum++;487
- if (m_bframeDelay)488
+ if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || !m_param->bDisableLookahead)489
{490
- int64_t *prevReorderedPts = m_prevReorderedPts;491
- frameEnc->m_dts = m_encodedFrameNum > m_bframeDelay492
- ? prevReorderedPts[(m_encodedFrameNum - m_bframeDelay) % m_bframeDelay]493
- : frameEnc->m_reorderedPts - m_bframeDelayTime;494
- prevReorderedPts[m_encodedFrameNum % m_bframeDelay] = frameEnc->m_reorderedPts;495
+ if (m_bframeDelay)496
+ {497
+ int64_t *prevReorderedPts = m_prevReorderedPts;498
+ frameEnc->m_dts = m_encodedFrameNum > m_bframeDelay499
+ ? prevReorderedPts[(m_encodedFrameNum - m_bframeDelay) % m_bframeDelay]500
+ : frameEnc->m_reorderedPts - m_bframeDelayTime;501
+ prevReorderedPts[m_encodedFrameNum % m_bframeDelay] = frameEnc->m_reorderedPts;502
+ }503
+ else504
+ frameEnc->m_dts = frameEnc->m_reorderedPts;505
}506
- else507
- frameEnc->m_dts = frameEnc->m_reorderedPts;508
509
/* Allocate analysis data before encode in save mode. This is allocated in frameEnc */510
if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE)511
512
513
uint32_t numCUsInFrame = widthInCU * heightInCU;514
analysis->numCUsInFrame = numCUsInFrame;515
+ analysis->numCuInHeight = heightInCU;516
analysis->numPartitions = m_param->num4x4Partitions;517
allocAnalysis(analysis);518
}519
520
521
int Encoder::reconfigureParam(x265_param* encParam, x265_param* param)522
{523
- encParam->maxNumReferences = param->maxNumReferences; // never uses more refs than specified in stream headers524
- encParam->bEnableFastIntra = param->bEnableFastIntra;525
- encParam->bEnableEarlySkip = param->bEnableEarlySkip;526
- encParam->bEnableRecursionSkip = param->bEnableRecursionSkip;527
- encParam->searchMethod = param->searchMethod;528
- /* Scratch buffer prevents me_range from being increased for esa/tesa */529
- if (param->searchRange < encParam->searchRange)530
- encParam->searchRange = param->searchRange;531
- /* We can't switch out of subme=0 during encoding. */532
- if (encParam->subpelRefine)533
- encParam->subpelRefine = param->subpelRefine;534
- encParam->rdoqLevel = param->rdoqLevel;535
- encParam->rdLevel = param->rdLevel;536
- encParam->bEnableRectInter = param->bEnableRectInter;537
- encParam->maxNumMergeCand = param->maxNumMergeCand;538
- encParam->bIntraInBFrames = param->bIntraInBFrames;539
- if (param->scalingLists && !encParam->scalingLists)540
- encParam->scalingLists = strdup(param->scalingLists);541
- /* VBV can't be turned ON if it wasn't ON to begin with and can't be turned OFF if it was ON to begin with*/542
- if (param->rc.vbvMaxBitrate > 0 && param->rc.vbvBufferSize > 0 &&543
- encParam->rc.vbvMaxBitrate > 0 && encParam->rc.vbvBufferSize > 0)544
- {545
- m_reconfigureRc |= encParam->rc.vbvMaxBitrate != param->rc.vbvMaxBitrate;546
- m_reconfigureRc |= encParam->rc.vbvBufferSize != param->rc.vbvBufferSize;547
- if (m_reconfigureRc && m_param->bEmitHRDSEI)548
- x265_log(m_param, X265_LOG_WARNING, "VBV parameters cannot be changed when HRD is in use.\n");549
- else550
- {551
- encParam->rc.vbvMaxBitrate = param->rc.vbvMaxBitrate;552
- encParam->rc.vbvBufferSize = param->rc.vbvBufferSize;553
+ if (isReconfigureRc(encParam, param))554
+ {555
+ /* VBV can't be turned ON if it wasn't ON to begin with and can't be turned OFF if it was ON to begin with*/556
+ if (param->rc.vbvMaxBitrate > 0 && param->rc.vbvBufferSize > 0 &&557
+ encParam->rc.vbvMaxBitrate > 0 && encParam->rc.vbvBufferSize > 0)558
+ {559
+ m_reconfigureRc |= encParam->rc.vbvMaxBitrate != param->rc.vbvMaxBitrate;560
+ m_reconfigureRc |= encParam->rc.vbvBufferSize != param->rc.vbvBufferSize;561
+ if (m_reconfigureRc && m_param->bEmitHRDSEI)562
+ x265_log(m_param, X265_LOG_WARNING, "VBV parameters cannot be changed when HRD is in use.\n");563
+ else564
+ {565
+ encParam->rc.vbvMaxBitrate = param->rc.vbvMaxBitrate;566
+ encParam->rc.vbvBufferSize = param->rc.vbvBufferSize;567
+ }568
}569
+ m_reconfigureRc |= encParam->rc.bitrate != param->rc.bitrate;570
+ encParam->rc.bitrate = param->rc.bitrate;571
+ m_reconfigureRc |= encParam->rc.rfConstant != param->rc.rfConstant;572
+ encParam->rc.rfConstant = param->rc.rfConstant;573
}574
- m_reconfigureRc |= encParam->rc.bitrate != param->rc.bitrate;575
- encParam->rc.bitrate = param->rc.bitrate;576
- m_reconfigureRc |= encParam->rc.rfConstant != param->rc.rfConstant;577
- encParam->rc.rfConstant = param->rc.rfConstant; 578
-579
+ else580
+ {581
+ encParam->maxNumReferences = param->maxNumReferences; // never uses more refs than specified in stream headers582
+ encParam->bEnableFastIntra = param->bEnableFastIntra;583
+ encParam->bEnableEarlySkip = param->bEnableEarlySkip;584
+ encParam->bEnableRecursionSkip = param->bEnableRecursionSkip;585
+ encParam->searchMethod = param->searchMethod;586
+ /* Scratch buffer prevents me_range from being increased for esa/tesa */587
+ if (param->searchRange < encParam->searchRange)588
+ encParam->searchRange = param->searchRange;589
+ /* We can't switch out of subme=0 during encoding. */590
+ if (encParam->subpelRefine)591
+ encParam->subpelRefine = param->subpelRefine;592
+ encParam->rdoqLevel = param->rdoqLevel;593
+ encParam->rdLevel = param->rdLevel;594
+ encParam->bEnableRectInter = param->bEnableRectInter;595
+ encParam->maxNumMergeCand = param->maxNumMergeCand;596
+ encParam->bIntraInBFrames = param->bIntraInBFrames;597
+ if (param->scalingLists && !encParam->scalingLists)598
+ encParam->scalingLists = strdup(param->scalingLists);599
+ }600
+ encParam->forceFlush = param->forceFlush;601
/* To add: Loop Filter/deblocking controls, transform skip, signhide require PPS to be resent */602
/* To add: SAO, temporal MVP, AMP, TU depths require SPS to be resent, at every CVS boundary */603
return x265_check_params(encParam);604
}605
606
+bool Encoder::isReconfigureRc(x265_param* latestParam, x265_param* param_in)607
+{608
+ return (latestParam->rc.vbvMaxBitrate != param_in->rc.vbvMaxBitrate609
+ || latestParam->rc.vbvBufferSize != param_in->rc.vbvBufferSize610
+ || latestParam->rc.bitrate != param_in->rc.bitrate611
+ || latestParam->rc.rfConstant != param_in->rc.rfConstant);612
+}613
+614
void Encoder::copyCtuInfo(x265_ctu_info_t** frameCtuInfo, int poc)615
{616
uint32_t widthInCU = (m_param->sourceWidth + m_param->maxCUSize - 1) >> m_param->maxLog2CUSize;617
618
void Encoder::configure(x265_param *p)619
{620
this->m_param = p;621
+ if (p->bMVType == AVC_INFO)622
+ this->m_externalFlush = true;623
+ else 624
+ this->m_externalFlush = false;625
if (p->keyframeMax < 0)626
{627
/* A negative max GOP size indicates the user wants only one I frame at628
629
x265_log(p, X265_LOG_WARNING, "MV refinement requires analysis load, analysis-reuse-level 10, scale factor. Disabling MV refine.\n");630
p->mvRefine = 0;631
}632
+ else if (p->interRefine >= 2)633
+ {634
+ x265_log(p, X265_LOG_WARNING, "MVs are recomputed when refine-inter >= 2. MV refinement not applicable. Disabling MV refine\n");635
+ p->mvRefine = 0;636
+ }637
}638
639
if ((p->analysisMultiPassRefine || p->analysisMultiPassDistortion) && (p->bDistributeModeAnalysis || p->bDistributeMotionEstimation))640
641
{642
X265_CHECK(analysis->sliceType, "invalid slice type\n");643
analysis->interData = analysis->intraData = NULL;644
+ if (m_param->bDisableLookahead)645
+ {646
+ CHECKED_MALLOC_ZERO(analysis->lookahead.intraSatdForVbv, uint32_t, analysis->numCuInHeight);647
+ CHECKED_MALLOC_ZERO(analysis->lookahead.satdForVbv, uint32_t, analysis->numCuInHeight);648
+ CHECKED_MALLOC_ZERO(analysis->lookahead.intraVbvCost, uint32_t, analysis->numCUsInFrame);649
+ CHECKED_MALLOC_ZERO(analysis->lookahead.vbvCost, uint32_t, analysis->numCUsInFrame);650
+ }651
if (analysis->sliceType == X265_TYPE_IDR || analysis->sliceType == X265_TYPE_I)652
{653
if (m_param->analysisReuseLevel < 2)654
655
{656
int numDir = analysis->sliceType == X265_TYPE_P ? 1 : 2;657
uint32_t numPlanes = m_param->internalCsp == X265_CSP_I400 ? 1 : 3;658
- CHECKED_MALLOC_ZERO(analysis->wt, WeightParam, numPlanes * numDir);659
+ if (!(m_param->bMVType == AVC_INFO))660
+ CHECKED_MALLOC_ZERO(analysis->wt, WeightParam, numPlanes * numDir);661
if (m_param->analysisReuseLevel < 2)662
return;663
664
665
CHECKED_MALLOC(interData->mergeFlag, uint8_t, analysis->numPartitions * analysis->numCUsInFrame);666
}667
668
- if (m_param->analysisReuseLevel == 10)669
+ if (m_param->analysisReuseLevel >= 7)670
{671
CHECKED_MALLOC(interData->interDir, uint8_t, analysis->numPartitions * analysis->numCUsInFrame);672
for (int dir = 0; dir < numDir; dir++)673
674
CHECKED_MALLOC(interData->mvpIdx[dir], uint8_t, analysis->numPartitions * analysis->numCUsInFrame);675
CHECKED_MALLOC(interData->refIdx[dir], int8_t, analysis->numPartitions * analysis->numCUsInFrame);676
CHECKED_MALLOC(interData->mv[dir], MV, analysis->numPartitions * analysis->numCUsInFrame);677
+ CHECKED_MALLOC(analysis->modeFlag[dir], uint8_t, analysis->numPartitions * analysis->numCUsInFrame);678
}679
680
/* Allocate intra in inter */681
682
683
void Encoder::freeAnalysis(x265_analysis_data* analysis)684
{685
+ if (m_param->bDisableLookahead)686
+ {687
+ X265_FREE(analysis->lookahead.satdForVbv);688
+ X265_FREE(analysis->lookahead.intraSatdForVbv);689
+ X265_FREE(analysis->lookahead.vbvCost);690
+ X265_FREE(analysis->lookahead.intraVbvCost);691
+ }692
/* Early exit freeing weights alone if level is 1 (when there is no analysis inter/intra) */693
- if (analysis->sliceType > X265_TYPE_I && analysis->wt)694
+ if (analysis->sliceType > X265_TYPE_I && analysis->wt && !(m_param->bMVType == AVC_INFO))695
X265_FREE(analysis->wt);696
if (m_param->analysisReuseLevel < 2)697
return;698
699
X265_FREE(((analysis_inter_data*)analysis->interData)->mergeFlag);700
X265_FREE(((analysis_inter_data*)analysis->interData)->partSize);701
}702
- if (m_param->analysisReuseLevel == 10)703
+ if (m_param->analysisReuseLevel >= 7)704
{705
X265_FREE(((analysis_inter_data*)analysis->interData)->interDir);706
+ X265_FREE(((analysis_inter_data*)analysis->interData)->sadCost);707
int numDir = analysis->sliceType == X265_TYPE_P ? 1 : 2;708
for (int dir = 0; dir < numDir; dir++)709
{710
X265_FREE(((analysis_inter_data*)analysis->interData)->mvpIdx[dir]);711
X265_FREE(((analysis_inter_data*)analysis->interData)->refIdx[dir]);712
X265_FREE(((analysis_inter_data*)analysis->interData)->mv[dir]);713
+ if (analysis->modeFlag[dir] != NULL)714
+ {715
+ X265_FREE(analysis->modeFlag[dir]);716
+ analysis->modeFlag[dir] = NULL;717
+ }718
}719
}720
else721
722
X265_FREAD(&analysis->satdCost, sizeof(int64_t), 1, m_analysisFile, &(picData->satdCost));723
X265_FREAD(&analysis->numCUsInFrame, sizeof(int), 1, m_analysisFile, &(picData->numCUsInFrame));724
X265_FREAD(&analysis->numPartitions, sizeof(int), 1, m_analysisFile, &(picData->numPartitions));725
+ if (m_param->bDisableLookahead)726
+ {727
+ X265_FREAD(&analysis->numCuInHeight, sizeof(uint32_t), 1, m_analysisFile, &(picData->numCuInHeight));728
+ X265_FREAD(&analysis->lookahead, sizeof(x265_lookahead_data), 1, m_analysisFile, &(picData->lookahead));729
+ }730
int scaledNumPartition = analysis->numPartitions;731
int factor = 1 << m_param->scaleFactor;732
733
734
735
/* Memory is allocated for inter and intra analysis data based on the slicetype */736
allocAnalysis(analysis);737
-738
+ if (m_param->bDisableLookahead)739
+ {740
+ X265_FREAD(analysis->lookahead.intraVbvCost, sizeof(uint32_t), analysis->numCUsInFrame, m_analysisFile, picData->lookahead.intraVbvCost);741
+ X265_FREAD(analysis->lookahead.vbvCost, sizeof(uint32_t), analysis->numCUsInFrame, m_analysisFile, picData->lookahead.vbvCost);742
+ X265_FREAD(analysis->lookahead.satdForVbv, sizeof(uint32_t), analysis->numCuInHeight, m_analysisFile, picData->lookahead.satdForVbv);743
+ X265_FREAD(analysis->lookahead.intraSatdForVbv, sizeof(uint32_t), analysis->numCuInHeight, m_analysisFile, picData->lookahead.intraSatdForVbv);744
+ }745
if (analysis->sliceType == X265_TYPE_IDR || analysis->sliceType == X265_TYPE_I)746
{747
if (m_param->analysisReuseLevel < 2)748
x265_2.5.tar.gz/source/encoder/encoder.h -> x265_2.6.tar.gz/source/encoder/encoder.h
Changed
39
1
2
RateControl* m_rateControl;3
Lookahead* m_lookahead;4
5
+ bool m_externalFlush;6
/* Collect statistics globally */7
EncStats m_analyzeAll;8
EncStats m_analyzeI;9
10
11
#ifdef ENABLE_HDR10_PLUS12
const hdr10plus_api *m_hdr10plus_api;13
- uint8_t **cim;14
- int numCimInfo;15
+ uint8_t **m_cim;16
+ int m_numCimInfo;17
#endif18
19
x265_sei_payload m_prevTonemapPayload;20
21
22
int reconfigureParam(x265_param* encParam, x265_param* param);23
24
+ bool isReconfigureRc(x265_param* latestParam, x265_param* param_in);25
+26
void copyCtuInfo(x265_ctu_info_t** frameCtuInfo, int poc);27
28
+ int copySlicetypePocAndSceneCut(int *slicetype, int *poc, int *sceneCut);29
+30
+ int getRefFrameList(PicYuv** l0, PicYuv** l1, int sliceType, int poc);31
+32
+ int setAnalysisDataAfterZScan(x265_analysis_data *analysis_data, Frame* curFrame);33
+34
+ int setAnalysisData(x265_analysis_data *analysis_data, int poc, uint32_t cuBytes);35
+36
void getStreamHeaders(NALList& list, Entropy& sbacCoder, Bitstream& bs);37
38
void fetchStats(x265_stats* stats, size_t statsSizeBytes);39
x265_2.5.tar.gz/source/encoder/frameencoder.cpp -> x265_2.6.tar.gz/source/encoder/frameencoder.cpp
Changed
579
1
2
delete[] m_outStreams;3
delete[] m_backupStreams;4
X265_FREE(m_sliceBaseRow);5
+ X265_FREE(m_sliceMaxBlockRow);6
X265_FREE(m_cuGeoms);7
X265_FREE(m_ctuGeomMap);8
X265_FREE(m_substreamSizes);9
10
11
m_sliceBaseRow = X265_MALLOC(uint32_t, m_param->maxSlices + 1);12
ok &= !!m_sliceBaseRow;13
+ m_sliceGroupSize = (uint16_t)(m_numRows + m_param->maxSlices - 1) / m_param->maxSlices;14
+ uint32_t sliceGroupSizeAccu = (m_numRows << 8) / m_param->maxSlices; 15
+ uint32_t rowSum = sliceGroupSizeAccu;16
+ uint32_t sidx = 0;17
+ for (uint32_t i = 0; i < m_numRows; i++)18
+ {19
+ const uint32_t rowRange = (rowSum >> 8);20
+ if ((i >= rowRange) & (sidx != m_param->maxSlices - 1))21
+ {22
+ rowSum += sliceGroupSizeAccu;23
+ m_sliceBaseRow[++sidx] = i;24
+ } 25
+ }26
+ X265_CHECK(sidx < m_param->maxSlices, "sliceID check failed!");27
+ m_sliceBaseRow[0] = 0;28
+ m_sliceBaseRow[m_param->maxSlices] = m_numRows;29
+30
+ m_sliceMaxBlockRow = X265_MALLOC(uint32_t, m_param->maxSlices + 1);31
+ ok &= !!m_sliceMaxBlockRow;32
+ uint32_t maxBlockRows = (m_param->sourceHeight + (16 - 1)) / 16;33
+ sliceGroupSizeAccu = (maxBlockRows << 8) / m_param->maxSlices;34
+ rowSum = sliceGroupSizeAccu;35
+ sidx = 0;36
+ for (uint32_t i = 0; i < maxBlockRows; i++)37
+ {38
+ const uint32_t rowRange = (rowSum >> 8);39
+ if ((i >= rowRange) & (sidx != m_param->maxSlices - 1))40
+ {41
+ rowSum += sliceGroupSizeAccu;42
+ m_sliceMaxBlockRow[++sidx] = i;43
+ }44
+ }45
+ m_sliceMaxBlockRow[0] = 0;46
+ m_sliceMaxBlockRow[m_param->maxSlices] = maxBlockRows;47
48
/* determine full motion search range */49
int range = m_param->searchRange; /* fpel search */50
51
while (!m_frame->m_ctuInfo)52
m_frame->m_copied.wait();53
}54
+ if ((m_param->bMVType == AVC_INFO) && !m_param->analysisReuseMode && !(IS_X265_TYPE_I(m_frame->m_lowres.sliceType)))55
+ {56
+ while (((m_frame->m_analysisData.interData == NULL && m_frame->m_analysisData.intraData == NULL) || (uint32_t)m_frame->m_poc != m_frame->m_analysisData.poc))57
+ m_frame->m_copyMVType.wait();58
+ }59
compressFrame();60
m_done.trigger(); /* FrameEncoder::getEncodedPicture() blocks for this event */61
+ if (m_frame != NULL)62
+ m_frame->m_reconEncoded.trigger();63
m_enable.wait();64
}65
}66
67
m_completionCount = 0;68
m_bAllRowsStop = false;69
m_vbvResetTriggerRow = -1;70
+ m_rowSliceTotalBits[0] = 0;71
+ m_rowSliceTotalBits[1] = 0;72
73
m_SSDY = m_SSDU = m_SSDV = 0;74
m_ssim = 0;75
76
77
/* reset entropy coders and compute slice id */78
m_entropyCoder.load(m_initSliceContext);79
- const uint32_t sliceGroupSize = (m_numRows + m_param->maxSlices - 1) / m_param->maxSlices;80
- const uint32_t sliceGroupSizeAccu = (m_numRows << 8) / m_param->maxSlices;81
- m_sliceGroupSize = (uint16_t)sliceGroupSize;82
+ 83
+ for (uint32_t sliceId = 0; sliceId < m_param->maxSlices; sliceId++) 84
+ for (uint32_t row = m_sliceBaseRow[sliceId]; row < m_sliceBaseRow[sliceId + 1]; row++)85
+ m_rows[row].init(m_initSliceContext, sliceId); 86
87
- uint32_t rowSum = sliceGroupSizeAccu;88
- uint32_t sidx = 0;89
- for (uint32_t i = 0; i < m_numRows; i++)90
- {91
- const uint32_t rowRange = (rowSum >> 8);92
-93
- if ((i >= rowRange) & (sidx != m_param->maxSlices - 1))94
- {95
- rowSum += sliceGroupSizeAccu;96
- m_sliceBaseRow[++sidx] = i;97
- }98
-99
- m_rows[i].init(m_initSliceContext, sidx);100
- }101
- X265_CHECK(sidx < m_param->maxSlices, "sliceID check failed!");102
-103
- m_sliceBaseRow[0] = 0;104
- m_sliceBaseRow[m_param->maxSlices] = m_numRows;105
+ // reset slice counter for rate control update106
+ m_sliceCnt = 0;107
108
uint32_t numSubstreams = m_param->bEnableWavefront ? slice->m_sps->numCuInHeight : m_param->maxSlices;109
X265_CHECK(m_param->bEnableWavefront || (m_param->maxSlices == 1), "Multiple slices without WPP unsupport now!");110
111
m_rows[i].rowGoOnCoder.setBitstream(&m_outStreams[i]);112
}113
else114
+ {115
for (uint32_t i = 0; i < numSubstreams; i++)116
m_outStreams[i].resetBits();117
+ }118
119
int prevBPSEI = m_rce.encodeOrder ? m_top->m_lastBPSEI : 0;120
121
122
* compressed in a wave-front pattern if WPP is enabled. Row based loop123
* filters runs behind the CTU compression and reconstruction */124
125
- for (uint32_t sliceId = 0; sliceId < m_param->maxSlices; sliceId++)126
- {127
+ for (uint32_t sliceId = 0; sliceId < m_param->maxSlices; sliceId++) 128
m_rows[m_sliceBaseRow[sliceId]].active = true;129
+ 130
+ if (m_param->bEnableWavefront)131
+ {132
+ int i = 0;133
+ for (uint32_t rowInSlice = 0; rowInSlice < m_sliceGroupSize; rowInSlice++)134
+ {135
+ for (uint32_t sliceId = 0; sliceId < m_param->maxSlices; sliceId++)136
+ {137
+ const uint32_t sliceStartRow = m_sliceBaseRow[sliceId];138
+ const uint32_t sliceEndRow = m_sliceBaseRow[sliceId + 1] - 1;139
+ const uint32_t row = sliceStartRow + rowInSlice;140
+ if (row > sliceEndRow)141
+ continue;142
+ m_row_to_idx[row] = i;143
+ m_idx_to_row[i] = row;144
+ i += 1;145
+ }146
+ }147
}148
149
if (m_param->bEnableWavefront)150
151
}152
}153
154
- enableRowEncoder(row); /* clear external dependency for this row */155
+ enableRowEncoder(m_row_to_idx[row]); /* clear external dependency for this row */156
if (!rowInSlice)157
{158
m_row0WaitTime = x265_mdate();159
- enqueueRowEncoder(row); /* clear internal dependency, start wavefront */160
+ enqueueRowEncoder(m_row_to_idx[row]); /* clear internal dependency, start wavefront */161
}162
tryWakeOne();163
} // end of loop rowInSlice164
165
// complete the slice header by writing WPP row-starts166
m_entropyCoder.setBitstream(&m_bs);167
if (slice->m_pps->bEntropyCodingSyncEnabled)168
- {169
m_entropyCoder.codeSliceHeaderWPPEntryPoints(&m_substreamSizes[prevSliceRow], (nextSliceRow - prevSliceRow - 1), maxStreamSize);170
- }171
+ 172
m_bs.writeByteAlignment();173
174
m_nalList.serialize(slice->m_nalUnitType, m_bs);175
176
if (ATOMIC_INC(&m_activeWorkerCount) == 1 && m_stallStartTime)177
m_totalNoWorkerTime += x265_mdate() - m_stallStartTime;178
179
- const uint32_t realRow = row >> 1;180
- const uint32_t typeNum = row & 1;181
+ const uint32_t realRow = m_idx_to_row[row >> 1];182
+ const uint32_t typeNum = m_idx_to_row[row & 1];183
184
if (!typeNum)185
processRowEncoder(realRow, m_tld[threadId]);186
187
188
// NOTE: Active next row189
if (realRow != m_sliceBaseRow[m_rows[realRow].sliceId + 1] - 1)190
- enqueueRowFilter(realRow + 1);191
+ enqueueRowFilter(m_row_to_idx[realRow + 1]);192
}193
194
if (ATOMIC_DEC(&m_activeWorkerCount) == 0)195
196
const uint32_t lineStartCUAddr = row * numCols;197
bool bIsVbv = m_param->rc.vbvBufferSize > 0 && m_param->rc.vbvMaxBitrate > 0;198
199
+ const uint32_t sliceId = curRow.sliceId;200
uint32_t maxBlockCols = (m_frame->m_fencPic->m_picWidth + (16 - 1)) / 16;201
- uint32_t maxBlockRows = (m_frame->m_fencPic->m_picHeight + (16 - 1)) / 16;202
uint32_t noOfBlocks = m_param->maxCUSize / 16;203
const uint32_t bFirstRowInSlice = ((row == 0) || (m_rows[row - 1].sliceId != curRow.sliceId)) ? 1 : 0;204
const uint32_t bLastRowInSlice = ((row == m_numRows - 1) || (m_rows[row + 1].sliceId != curRow.sliceId)) ? 1 : 0;205
- const uint32_t sliceId = curRow.sliceId;206
const uint32_t endRowInSlicePlus1 = m_sliceBaseRow[sliceId + 1];207
const uint32_t rowInSlice = row - m_sliceBaseRow[sliceId];208
209
- if (bFirstRowInSlice && !curRow.completed)210
- {211
- // Load SBAC coder context from previous row and initialize row state.212
- //rowCoder.copyState(m_initSliceContext);213
- //rowCoder.loadContexts(m_rows[row - 1].bufferedEntropy);214
- rowCoder.load(m_initSliceContext);215
- //m_rows[row - 1].bufferedEntropy.loadContexts(m_initSliceContext);216
- }217
+ // Load SBAC coder context from previous row and initialize row state.218
+ if (bFirstRowInSlice && !curRow.completed) 219
+ rowCoder.load(m_initSliceContext); 220
221
// calculate mean QP for consistent deltaQP signalling calculation222
if (m_param->bOptCUDeltaQP)223
224
if (m_param->bEnableWavefront || !row)225
{226
double meanQPOff = 0;227
- uint32_t loopIncr, count = 0;228
bool isReferenced = IS_REFERENCED(m_frame);229
double *qpoffs = (isReferenced && m_param->rc.cuTree) ? m_frame->m_lowres.qpCuTreeOffset : m_frame->m_lowres.qpAqOffset;230
if (qpoffs)231
{232
- if (m_param->rc.qgSize == 8)233
- loopIncr = 8;234
- else235
- loopIncr = 16;236
+ uint32_t loopIncr = (m_param->rc.qgSize == 8) ? 8 : 16;237
+238
uint32_t cuYStart = 0, height = m_frame->m_fencPic->m_picHeight;239
if (m_param->bEnableWavefront)240
{241
242
243
uint32_t qgSize = m_param->rc.qgSize, width = m_frame->m_fencPic->m_picWidth;244
uint32_t maxOffsetCols = (m_frame->m_fencPic->m_picWidth + (loopIncr - 1)) / loopIncr;245
+ uint32_t count = 0;246
for (uint32_t cuY = cuYStart; cuY < height && (cuY < m_frame->m_fencPic->m_picHeight); cuY += qgSize)247
{248
for (uint32_t cuX = 0; cuX < width; cuX += qgSize)249
250
}251
curRow.avgQPComputed = 1;252
}253
- }254
-255
- // TODO: specially case handle on first and last row256
+ } 257
258
// Initialize restrict on MV range in slices259
tld.analysis.m_sliceMinY = -(int16_t)(rowInSlice * m_param->maxCUSize * 4) + 3 * 4;260
261
curRow.bufferedEntropy.copyState(rowCoder);262
curRow.bufferedEntropy.loadContexts(rowCoder);263
}264
- if (!row && m_vbvResetTriggerRow != intRow)265
+ if (bFirstRowInSlice && m_vbvResetTriggerRow != intRow) 266
{267
curEncData.m_rowStat[row].rowQp = curEncData.m_avgQpRc;268
curEncData.m_rowStat[row].rowQpScale = x265_qp2qScale(curEncData.m_avgQpRc);269
}270
271
FrameData::RCStatCU& cuStat = curEncData.m_cuStat[cuAddr];272
- if (m_param->bEnableWavefront && row >= col && row && m_vbvResetTriggerRow != intRow)273
+ if (m_param->bEnableWavefront && rowInSlice >= col && !bFirstRowInSlice && m_vbvResetTriggerRow != intRow)274
cuStat.baseQp = curEncData.m_cuStat[cuAddr - numCols + 1].baseQp;275
- else if (!m_param->bEnableWavefront && row && m_vbvResetTriggerRow != intRow)276
+ else if (!m_param->bEnableWavefront && !bFirstRowInSlice && m_vbvResetTriggerRow != intRow)277
cuStat.baseQp = curEncData.m_rowStat[row - 1].rowQp;278
else279
cuStat.baseQp = curEncData.m_rowStat[row].rowQp;280
281
/* TODO: use defines from slicetype.h for lowres block size */282
uint32_t block_y = (ctu->m_cuPelY >> m_param->maxLog2CUSize) * noOfBlocks;283
uint32_t block_x = (ctu->m_cuPelX >> m_param->maxLog2CUSize) * noOfBlocks;284
- 285
- cuStat.vbvCost = 0;286
- cuStat.intraVbvCost = 0;287
- for (uint32_t h = 0; h < noOfBlocks && block_y < maxBlockRows; h++, block_y++)288
+ if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || !m_param->bDisableLookahead)289
{290
- uint32_t idx = block_x + (block_y * maxBlockCols);291
+ cuStat.vbvCost = 0;292
+ cuStat.intraVbvCost = 0;293
294
- for (uint32_t w = 0; w < noOfBlocks && (block_x + w) < maxBlockCols; w++, idx++)295
+ for (uint32_t h = 0; h < noOfBlocks && block_y < m_sliceMaxBlockRow[sliceId + 1]; h++, block_y++)296
{297
- cuStat.vbvCost += m_frame->m_lowres.lowresCostForRc[idx] & LOWRES_COST_MASK;298
- cuStat.intraVbvCost += m_frame->m_lowres.intraCost[idx];299
+ uint32_t idx = block_x + (block_y * maxBlockCols);300
+301
+ for (uint32_t w = 0; w < noOfBlocks && (block_x + w) < maxBlockCols; w++, idx++)302
+ {303
+ cuStat.vbvCost += m_frame->m_lowres.lowresCostForRc[idx] & LOWRES_COST_MASK;304
+ cuStat.intraVbvCost += m_frame->m_lowres.intraCost[idx];305
+ }306
}307
}308
}309
310
{311
// NOTE: in VBV mode, we may reencode anytime, so we can't do Deblock stage-Horizon and SAO312
if (!bIsVbv)313
- {314
- // TODO: Multiple Threading315
- // Delay ONE row to avoid Intra Prediction Conflict316
+ { 317
+ // Delay one row to avoid intra prediction conflict318
if (m_pool && !bFirstRowInSlice)319
- {320
- // Waitting last threading finish321
- m_frameFilter.m_parallelFilter[row - 1].waitForExit();322
-323
- // Processing new group324
+ { 325
int allowCol = col;326
327
// avoid race condition on last column328
329
: m_frameFilter.m_parallelFilter[row - 2].m_lastCol.get()), (int)col);330
}331
m_frameFilter.m_parallelFilter[row - 1].m_allowedCol.set(allowCol);332
- m_frameFilter.m_parallelFilter[row - 1].tryBondPeers(*this, 1);333
}334
335
// Last Row may start early336
if (m_pool && bLastRowInSlice)337
{338
- // Waiting for the last thread to finish339
- m_frameFilter.m_parallelFilter[row].waitForExit();340
-341
// Deblocking last row342
int allowCol = col;343
344
345
: m_frameFilter.m_parallelFilter[row - 1].m_lastCol.get()), (int)col);346
}347
m_frameFilter.m_parallelFilter[row].m_allowedCol.set(allowCol);348
- m_frameFilter.m_parallelFilter[row].tryBondPeers(*this, 1);349
}350
} // end of !bIsVbv351
}352
353
FrameStats frameLog;354
curEncData.m_rowStat[row].sumQpAq += collectCTUStatistics(*ctu, &frameLog);355
356
- // copy no. of intra, inter Cu cnt per row into frame stats for 2 pass357
+ // copy number of intra, inter cu per row into frame stats for 2 pass358
if (m_param->rc.bStatWrite)359
{360
curRow.rowStats.mvBits += best.mvBits;361
362
int shift = 2 * (m_param->maxCUDepth - depth);363
int cuSize = m_param->maxCUSize >> depth;364
365
- if (cuSize == 8)366
- curRow.rowStats.intra8x8Cnt += (int)(frameLog.cntIntra[depth] + frameLog.cntIntraNxN);367
- else368
- curRow.rowStats.intra8x8Cnt += (int)(frameLog.cntIntra[depth] << shift);369
+ curRow.rowStats.intra8x8Cnt += (cuSize == 8) ? (int)(frameLog.cntIntra[depth] + frameLog.cntIntraNxN) :370
+ (int)(frameLog.cntIntra[depth] << shift);371
372
curRow.rowStats.inter8x8Cnt += (int)(frameLog.cntInter[depth] << shift);373
curRow.rowStats.skip8x8Cnt += (int)((frameLog.cntSkipCu[depth] + frameLog.cntMergeCu[depth]) << shift);374
375
if (bIsVbv)376
{ 377
// Update encoded bits, satdCost, baseQP for each CU if tune grain is disabled378
- if ((m_param->bEnableWavefront && (!cuAddr || !m_param->rc.bEnableConstVbv)) || !m_param->bEnableWavefront)379
+ FrameData::RCStatCU& cuStat = curEncData.m_cuStat[cuAddr]; 380
+ if ((m_param->bEnableWavefront && ((cuAddr == m_sliceBaseRow[sliceId] * numCols) || !m_param->rc.bEnableConstVbv)) || !m_param->bEnableWavefront)381
{382
- curEncData.m_rowStat[row].rowSatd += curEncData.m_cuStat[cuAddr].vbvCost;383
- curEncData.m_rowStat[row].rowIntraSatd += curEncData.m_cuStat[cuAddr].intraVbvCost;384
- curEncData.m_rowStat[row].encodedBits += curEncData.m_cuStat[cuAddr].totalBits;385
- curEncData.m_rowStat[row].sumQpRc += curEncData.m_cuStat[cuAddr].baseQp;386
+ curEncData.m_rowStat[row].rowSatd += cuStat.vbvCost;387
+ curEncData.m_rowStat[row].rowIntraSatd += cuStat.intraVbvCost;388
+ curEncData.m_rowStat[row].encodedBits += cuStat.totalBits;389
+ curEncData.m_rowStat[row].sumQpRc += cuStat.baseQp;390
curEncData.m_rowStat[row].numEncodedCUs = cuAddr;391
}392
393
// If current block is at row end checkpoint, call vbv ratecontrol.394
-395
if (!m_param->bEnableWavefront && col == numCols - 1)396
{397
double qpBase = curEncData.m_cuStat[cuAddr].baseQp;398
- int reEncode = m_top->m_rateControl->rowVbvRateControl(m_frame, row, &m_rce, qpBase);399
+ int reEncode = m_top->m_rateControl->rowVbvRateControl(m_frame, row, &m_rce, qpBase, m_sliceBaseRow, sliceId);400
qpBase = x265_clip3((double)m_param->rc.qpMin, (double)m_param->rc.qpMax, qpBase);401
curEncData.m_rowStat[row].rowQp = qpBase;402
curEncData.m_rowStat[row].rowQpScale = x265_qp2qScale(qpBase);403
404
curEncData.m_rowStat[row].sumQpAq = 0;405
}406
}407
-408
// If current block is at row diagonal checkpoint, call vbv ratecontrol.409
-410
- else if (m_param->bEnableWavefront && row == col && row)411
+ else if (m_param->bEnableWavefront && rowInSlice == col && !bFirstRowInSlice)412
{413
if (m_param->rc.bEnableConstVbv)414
{415
- int32_t startCuAddr = numCols * row;416
- int32_t EndCuAddr = startCuAddr + col;417
- for (int32_t r = row; r >= 0; r--)418
+ uint32_t startCuAddr = numCols * row;419
+ uint32_t EndCuAddr = startCuAddr + col;420
+421
+ for (int32_t r = row; r >= (int32_t)m_sliceBaseRow[sliceId]; r--)422
{423
- for (int32_t c = startCuAddr; c <= EndCuAddr && c <= (int32_t)numCols * (r + 1) - 1; c++)424
+ for (uint32_t c = startCuAddr; c <= EndCuAddr && c <= numCols * (r + 1) - 1; c++)425
{426
curEncData.m_rowStat[r].rowSatd += curEncData.m_cuStat[c].vbvCost;427
curEncData.m_rowStat[r].rowIntraSatd += curEncData.m_cuStat[c].intraVbvCost;428
429
}430
}431
double qpBase = curEncData.m_cuStat[cuAddr].baseQp;432
- int reEncode = m_top->m_rateControl->rowVbvRateControl(m_frame, row, &m_rce, qpBase);433
+ int reEncode = m_top->m_rateControl->rowVbvRateControl(m_frame, row, &m_rce, qpBase, m_sliceBaseRow, sliceId);434
qpBase = x265_clip3((double)m_param->rc.qpMin, (double)m_param->rc.qpMax, qpBase);435
curEncData.m_rowStat[row].rowQp = qpBase;436
- curEncData.m_rowStat[row].rowQpScale = x265_qp2qScale(qpBase);437
+ curEncData.m_rowStat[row].rowQpScale = x265_qp2qScale(qpBase);438
439
if (reEncode < 0)440
{441
442
m_vbvResetTriggerRow = row;443
m_bAllRowsStop = true;444
445
- for (uint32_t r = m_numRows - 1; r >= row; r--)446
+ for (uint32_t r = m_sliceBaseRow[sliceId + 1] - 1; r >= row; r--)447
{448
CTURow& stopRow = m_rows[r];449
450
451
m_rows[row + 1].completed + 2 <= curRow.completed)452
{453
m_rows[row + 1].active = true;454
- enqueueRowEncoder(row + 1);455
+ enqueueRowEncoder(m_row_to_idx[row + 1]);456
tryWakeOne(); /* wake up a sleeping thread or set the help wanted flag */457
}458
}459
460
}461
}462
463
- /** this row of CTUs has been compressed **/464
+ /* this row of CTUs has been compressed */465
if (m_param->bEnableWavefront && m_param->rc.bEnableConstVbv)466
{467
- if (row == m_numRows - 1)468
+ if (bLastRowInSlice) 469
{470
- for (int32_t r = 0; r < (int32_t)m_numRows; r++)471
+ for (uint32_t r = m_sliceBaseRow[sliceId]; r < m_sliceBaseRow[sliceId + 1]; r++)472
{473
- for (int32_t c = curEncData.m_rowStat[r].numEncodedCUs + 1; c < (int32_t)numCols * (r + 1); c++)474
+ for (uint32_t c = curEncData.m_rowStat[r].numEncodedCUs + 1; c < numCols * (r + 1); c++)475
{476
curEncData.m_rowStat[r].rowSatd += curEncData.m_cuStat[c].vbvCost;477
curEncData.m_rowStat[r].rowIntraSatd += curEncData.m_cuStat[c].intraVbvCost;478
479
* after half the frame is encoded, but after this initial period we update480
* after refLagRows (the number of rows reference frames must have completed481
* before referencees may begin encoding) */482
- uint32_t rowCount = 0;483
if (m_param->rc.rateControlMode == X265_RC_ABR || bIsVbv)484
{485
+ uint32_t rowCount = 0;486
+ uint32_t maxRows = m_sliceBaseRow[sliceId + 1] - m_sliceBaseRow[sliceId];487
+488
if (!m_rce.encodeOrder)489
- rowCount = m_numRows - 1;490
+ rowCount = maxRows - 1; 491
else if ((uint32_t)m_rce.encodeOrder <= 2 * (m_param->fpsNum / m_param->fpsDenom))492
- rowCount = X265_MIN((m_numRows + 1) / 2, m_numRows - 1);493
+ rowCount = X265_MIN((maxRows + 1) / 2, maxRows - 1);494
else495
- rowCount = X265_MIN(m_refLagRows, m_numRows - 1);496
- if (row == rowCount)497
+ rowCount = X265_MIN(m_refLagRows / m_param->maxSlices, maxRows - 1);498
+499
+ if (rowInSlice == rowCount)500
{501
- m_rce.rowTotalBits = 0;502
+ m_rowSliceTotalBits[sliceId] = 0;503
if (bIsVbv)504
- for (uint32_t i = 0; i < rowCount; i++)505
- m_rce.rowTotalBits += curEncData.m_rowStat[i].encodedBits;506
+ { 507
+ for (uint32_t i = m_sliceBaseRow[sliceId]; i < rowCount + m_sliceBaseRow[sliceId]; i++)508
+ m_rowSliceTotalBits[sliceId] += curEncData.m_rowStat[i].encodedBits;509
+ }510
else511
- for (uint32_t cuAddr = 0; cuAddr < rowCount * numCols; cuAddr++)512
- m_rce.rowTotalBits += curEncData.m_cuStat[cuAddr].totalBits;513
-514
- m_top->m_rateControl->rateControlUpdateStats(&m_rce);515
+ {516
+ uint32_t startAddr = m_sliceBaseRow[sliceId] * numCols;517
+ uint32_t finishAddr = startAddr + rowCount * numCols;518
+ 519
+ for (uint32_t cuAddr = startAddr; cuAddr < finishAddr; cuAddr++)520
+ m_rowSliceTotalBits[sliceId] += curEncData.m_cuStat[cuAddr].totalBits;521
+ } 522
+523
+ if (ATOMIC_INC(&m_sliceCnt) == (int)m_param->maxSlices)524
+ {525
+ m_rce.rowTotalBits = 0;526
+ for (uint32_t i = 0; i < m_param->maxSlices; i++)527
+ m_rce.rowTotalBits += m_rowSliceTotalBits[i];528
+ m_top->m_rateControl->rateControlUpdateStats(&m_rce);529
+ }530
}531
}532
533
534
/* Processing left Deblock block with current threading */535
if ((m_param->bEnableLoopFilter | m_param->bEnableSAO) & (rowInSlice >= 2))536
{537
- /* TODO: Multiple Threading */538
-539
/* Check conditional to start previous row process with current threading */540
if (m_frameFilter.m_parallelFilter[row - 2].m_lastDeblocked.get() == (int)numCols)541
{542
/* stop threading on current row and restart it */543
- m_frameFilter.m_parallelFilter[row - 1].waitForExit();544
m_frameFilter.m_parallelFilter[row - 1].m_allowedCol.set(numCols);545
m_frameFilter.m_parallelFilter[row - 1].processTasks(-1);546
}547
548
{549
if (rowInSlice >= m_filterRowDelay)550
{551
- enableRowFilter(row - m_filterRowDelay);552
+ enableRowFilter(m_row_to_idx[row - m_filterRowDelay]);553
554
/* NOTE: Activate filter if first row (row 0) */555
if (rowInSlice == m_filterRowDelay)556
- enqueueRowFilter(row - m_filterRowDelay);557
+ enqueueRowFilter(m_row_to_idx[row - m_filterRowDelay]);558
tryWakeOne();559
}560
561
562
{563
for (uint32_t i = endRowInSlicePlus1 - m_filterRowDelay; i < endRowInSlicePlus1; i++)564
{565
- enableRowFilter(i);566
+ enableRowFilter(m_row_to_idx[i]);567
}568
tryWakeOne();569
}570
571
// handle specially case - single row slice572
if (bFirstRowInSlice & bLastRowInSlice)573
{574
- enqueueRowFilter(row);575
+ enqueueRowFilter(m_row_to_idx[row]);576
tryWakeOne();577
}578
}579
x265_2.5.tar.gz/source/encoder/frameencoder.h -> x265_2.6.tar.gz/source/encoder/frameencoder.h
Changed
22
1
2
volatile bool m_bAllRowsStop;3
volatile int m_completionCount;4
volatile int m_vbvResetTriggerRow;5
+ volatile int m_sliceCnt;6
7
uint32_t m_numRows;8
uint32_t m_numCols;9
10
11
CTURow* m_rows;12
uint16_t m_sliceAddrBits;13
- uint16_t m_sliceGroupSize;14
- uint32_t* m_sliceBaseRow;15
+ uint32_t m_sliceGroupSize;16
+ uint32_t* m_sliceBaseRow; 17
+ uint32_t* m_sliceMaxBlockRow;18
+ int64_t m_rowSliceTotalBits[2];19
RateControlEntry m_rce;20
SEIDecodedPictureHash m_seiReconPictureDigest;21
22
x265_2.5.tar.gz/source/encoder/framefilter.cpp -> x265_2.6.tar.gz/source/encoder/framefilter.cpp
Changed
21
1
2
CUData* ctu = encData.getPicCTU(m_parallelFilter[row].m_rowAddr);3
4
/* Processing left block Deblock with current threading */5
- {6
- /* stop threading on current row */7
- m_parallelFilter[row].waitForExit();8
-9
+ { 10
/* Check to avoid previous row process slower than current row */11
X265_CHECK(ctu->m_bFirstRowInSlice || m_parallelFilter[row - 1].m_lastDeblocked.get() == m_numCols, "previous row not finish");12
13
14
}15
16
// this row of CTUs has been encoded17
-18
if (!ctu->m_bFirstRowInSlice)19
processPostRow(row - 1);20
21
x265_2.5.tar.gz/source/encoder/framefilter.h -> x265_2.6.tar.gz/source/encoder/framefilter.h
Changed
21
1
2
void* m_ssimBuf; /* Temp storage for ssim computation */3
4
#define MAX_PFILTER_CUS (4) /* maximum CUs for every thread */5
- class ParallelFilter : public BondedTaskGroup, public Deblock6
+ class ParallelFilter : public Deblock7
{8
public:9
uint32_t m_rowHeight;10
11
{12
return m_rowHeight;13
}14
-15
- protected:16
-17
- ParallelFilter operator=(const ParallelFilter&);18
};19
20
ParallelFilter* m_parallelFilter;21
x265_2.5.tar.gz/source/encoder/ratecontrol.cpp -> x265_2.6.tar.gz/source/encoder/ratecontrol.cpp
Changed
253
1
2
m_param->rc.vbvBufferSize = x265_clip3(0, 2000000, m_param->rc.vbvBufferSize);3
m_param->rc.vbvMaxBitrate = x265_clip3(0, 2000000, m_param->rc.vbvMaxBitrate);4
m_param->rc.vbvBufferInit = x265_clip3(0.0, 2000000.0, m_param->rc.vbvBufferInit);5
+ m_param->vbvBufferEnd = x265_clip3(0.0, 2000000.0, m_param->vbvBufferEnd);6
m_singleFrameVbv = 0;7
m_rateTolerance = 1.0;8
9
10
m_param->rc.vbvMaxBitrate = 0;11
}12
m_isVbv = m_param->rc.vbvMaxBitrate > 0 && m_param->rc.vbvBufferSize > 0;13
+ if (m_param->vbvBufferEnd && !m_isVbv)14
+ {15
+ x265_log(m_param, X265_LOG_WARNING, "vbv-end requires VBV parameters, ignored\n");16
+ m_param->vbvBufferEnd = 0;17
+ }18
if (m_param->bEmitHRDSEI && !m_isVbv)19
{20
x265_log(m_param, X265_LOG_WARNING, "NAL HRD parameters require VBV parameters, ignored\n");21
22
23
if (m_param->rc.vbvBufferInit > 1.)24
m_param->rc.vbvBufferInit = x265_clip3(0.0, 1.0, m_param->rc.vbvBufferInit / m_param->rc.vbvBufferSize);25
+ if (m_param->vbvBufferEnd > 1.)26
+ m_param->vbvBufferEnd = x265_clip3(0.0, 1.0, m_param->vbvBufferEnd / m_param->rc.vbvBufferSize);27
+ if (m_param->vbvEndFrameAdjust > 1.)28
+ m_param->vbvEndFrameAdjust = x265_clip3(0.0, 1.0, m_param->vbvEndFrameAdjust);29
m_param->rc.vbvBufferInit = x265_clip3(0.0, 1.0, X265_MAX(m_param->rc.vbvBufferInit, m_bufferRate / m_bufferSize));30
m_bufferFillFinal = m_bufferSize * m_param->rc.vbvBufferInit;31
m_bufferFillActual = m_bufferFillFinal;32
33
m_bitrate = m_param->rc.bitrate * 1000;34
}35
36
-37
void RateControl::initHRD(SPS& sps)38
{39
int vbvBufferSize = m_param->rc.vbvBufferSize * 1000;40
41
42
#undef MAX_DURATION43
}44
+45
bool RateControl::analyseABR2Pass(uint64_t allAvailableBits)46
{47
double rateFactor, stepMult;48
49
50
return q;51
}52
+53
double RateControl::countExpectedBits(int startPos, int endPos)54
{55
double expectedBits = 0;56
57
}58
return expectedBits;59
}60
+61
bool RateControl::findUnderflow(double *fills, int *t0, int *t1, int over, int endPos)62
{63
/* find an interval ending on an overflow or underflow (depending on whether64
65
}66
return adjusted;67
}68
+69
bool RateControl::cuTreeReadFor2Pass(Frame* frame)70
{71
int index = m_encOrder[frame->m_poc];72
73
double RateControl::tuneAbrQScaleFromFeedback(double qScale)74
{75
double abrBuffer = 2 * m_rateTolerance * m_bitrate;76
- /* use framesDone instead of POC as poc count is not serial with bframes enabled */77
- double overflow = 1.0;78
- double timeDone = (double)(m_framesDone - m_param->frameNumThreads + 1) * m_frameDuration;79
- double wantedBits = timeDone * m_bitrate;80
- int64_t encodedBits = m_totalBits;81
- if (m_param->totalFrames && m_param->totalFrames <= 2 * m_fps)82
- {83
- abrBuffer = m_param->totalFrames * (m_bitrate / m_fps);84
- encodedBits = m_encodedBits;85
- }86
-87
- if (wantedBits > 0 && encodedBits > 0 && (!m_partialResidualFrames || 88
- m_param->rc.bStrictCbr || m_isGrainEnabled))89
- {90
- abrBuffer *= X265_MAX(1, sqrt(timeDone));91
- overflow = x265_clip3(.5, 2.0, 1.0 + (encodedBits - wantedBits) / abrBuffer);92
- qScale *= overflow;93
- }94
+ /* use framesDone instead of POC as poc count is not serial with bframes enabled */95
+ double overflow = 1.0;96
+ double timeDone = (double)(m_framesDone - m_param->frameNumThreads + 1) * m_frameDuration;97
+ double wantedBits = timeDone * m_bitrate;98
+ int64_t encodedBits = m_totalBits;99
+ if (m_param->totalFrames && m_param->totalFrames <= 2 * m_fps)100
+ {101
+ abrBuffer = m_param->totalFrames * (m_bitrate / m_fps);102
+ encodedBits = m_encodedBits;103
+ }104
+105
+ if (wantedBits > 0 && encodedBits > 0 && (!m_partialResidualFrames || 106
+ m_param->rc.bStrictCbr || m_isGrainEnabled))107
+ {108
+ abrBuffer *= X265_MAX(1, sqrt(timeDone));109
+ overflow = x265_clip3(.5, 2.0, 1.0 + (encodedBits - wantedBits) / abrBuffer);110
+ qScale *= overflow;111
+ }112
return qScale;113
}114
115
116
curBits = predictSize(&m_pred[predType], frameQ[type], (double)satd);117
bufferFillCur -= curBits;118
}119
-120
- /* Try to get the buffer at least 50% filled, but don't set an impossible goal. */121
- double finalDur = 1;122
- if (m_param->rc.bStrictCbr)123
- {124
- finalDur = x265_clip3(0.4, 1.0, totalDuration);125
- }126
- targetFill = X265_MIN(m_bufferFill + totalDuration * m_vbvMaxRate * 0.5 , m_bufferSize * (1 - 0.5 * finalDur));127
- if (bufferFillCur < targetFill)128
- {129
- q *= 1.01;130
- loopTerminate |= 1;131
- continue;132
- }133
- /* Try to get the buffer not more than 80% filled, but don't set an impossible goal. */134
- targetFill = x265_clip3(m_bufferSize * (1 - 0.2 * finalDur), m_bufferSize, m_bufferFill - totalDuration * m_vbvMaxRate * 0.5);135
- if (m_isCbr && bufferFillCur > targetFill && !m_isSceneTransition)136
- {137
- q /= 1.01;138
- loopTerminate |= 2;139
- continue;140
+ if (m_param->vbvBufferEnd && rce->encodeOrder >= m_param->vbvEndFrameAdjust * m_param->totalFrames)141
+ {142
+ bool loopBreak = false;143
+ double bufferDiff = m_param->vbvBufferEnd - (m_bufferFill / m_bufferSize);144
+ targetFill = m_bufferFill + m_bufferSize * (bufferDiff / (m_param->totalFrames - rce->encodeOrder));145
+ if (bufferFillCur < targetFill)146
+ {147
+ q *= 1.01;148
+ loopTerminate |= 1;149
+ loopBreak = true;150
+ }151
+ if (bufferFillCur > m_param->vbvBufferEnd * m_bufferSize)152
+ {153
+ q /= 1.01;154
+ loopTerminate |= 2;155
+ loopBreak = true;156
+ }157
+ if (!loopBreak)158
+ break;159
+ }160
+ else161
+ {162
+ /* Try to get the buffer at least 50% filled, but don't set an impossible goal. */163
+ double finalDur = 1;164
+ if (m_param->rc.bStrictCbr)165
+ {166
+ finalDur = x265_clip3(0.4, 1.0, totalDuration);167
+ }168
+ targetFill = X265_MIN(m_bufferFill + totalDuration * m_vbvMaxRate * 0.5, m_bufferSize * (1 - 0.5 * finalDur));169
+ if (bufferFillCur < targetFill)170
+ {171
+ q *= 1.01;172
+ loopTerminate |= 1;173
+ continue;174
+ }175
+ /* Try to get the buffer not more than 80% filled, but don't set an impossible goal. */176
+ targetFill = x265_clip3(m_bufferSize * (1 - 0.2 * finalDur), m_bufferSize, m_bufferFill - totalDuration * m_vbvMaxRate * 0.5);177
+ if (m_isCbr && bufferFillCur > targetFill && !m_isSceneTransition)178
+ {179
+ q /= 1.01;180
+ loopTerminate |= 2;181
+ continue;182
+ }183
+ break;184
}185
- break;186
}187
q = X265_MAX(q0 / 2, q);188
}189
190
return totalSatdBits + encodedBitsSoFar;191
}192
193
-int RateControl::rowVbvRateControl(Frame* curFrame, uint32_t row, RateControlEntry* rce, double& qpVbv)194
+int RateControl::rowVbvRateControl(Frame* curFrame, uint32_t row, RateControlEntry* rce, double& qpVbv, uint32_t* m_sliceBaseRow, uint32_t sliceId)195
{196
FrameData& curEncData = *curFrame->m_encData;197
double qScaleVbv = x265_qp2qScale(qpVbv);198
uint64_t rowSatdCost = curEncData.m_rowStat[row].rowSatd;199
double encodedBits = curEncData.m_rowStat[row].encodedBits;200
+ uint32_t rowInSlice = row - m_sliceBaseRow[sliceId];201
202
- if (m_param->bEnableWavefront && row == 1)203
+ if (m_param->bEnableWavefront && rowInSlice == 1)204
{205
- rowSatdCost += curEncData.m_rowStat[0].rowSatd;206
- encodedBits += curEncData.m_rowStat[0].encodedBits;207
+ rowSatdCost += curEncData.m_rowStat[row - 1].rowSatd;208
+ encodedBits += curEncData.m_rowStat[row - 1].encodedBits;209
}210
rowSatdCost >>= X265_DEPTH - 8;211
updatePredictor(rce->rowPred[0], qScaleVbv, (double)rowSatdCost, encodedBits);212
213
if (qpVbv < refFrame->m_encData->m_rowStat[row].rowQp)214
{215
uint64_t intraRowSatdCost = curEncData.m_rowStat[row].rowIntraSatd;216
- if (m_param->bEnableWavefront && row == 1)217
- intraRowSatdCost += curEncData.m_rowStat[0].rowIntraSatd;218
+ if (m_param->bEnableWavefront && rowInSlice == 1)219
+ intraRowSatdCost += curEncData.m_rowStat[row - 1].rowIntraSatd;220
intraRowSatdCost >>= X265_DEPTH - 8;221
updatePredictor(rce->rowPred[1], qScaleVbv, (double)intraRowSatdCost, encodedBits);222
}223
224
const SPS& sps = *curEncData.m_slice->m_sps;225
double maxFrameError = X265_MAX(0.05, 1.0 / sps.numCuInHeight);226
227
- if (row < sps.numCuInHeight - 1)228
+ if (row < m_sliceBaseRow[sliceId + 1] - 1)229
{230
/* More threads means we have to be more cautious in letting ratecontrol use up extra bits. */231
double rcTol = bufferLeftPlanned / m_param->frameNumThreads * m_rateTolerance;232
233
m_encodedBitsWindow[pos % s_slidingWindowFrames] = actualBits;234
if(rce->sliceType != I_SLICE)235
{236
- int qp = int (rce->qpaRc + 0.5);237
- m_qpToEncodedBits[qp] = m_qpToEncodedBits[qp] == 0 ? actualBits : (m_qpToEncodedBits[qp] + actualBits) * 0.5;238
+ int qp = int (rce->qpaRc + 0.5);239
+ m_qpToEncodedBits[qp] = m_qpToEncodedBits[qp] == 0 ? actualBits : (m_qpToEncodedBits[qp] + actualBits) * 0.5;240
}241
curFrame->m_rcData->wantedBitsWindow = m_wantedBitsWindow;242
curFrame->m_rcData->cplxrSum = m_cplxrSum;243
244
curFrame->m_encData->m_frameStats.percent8x8Skip * m_ncu) < 0)245
goto writeFailure;246
}247
- else{248
+ else249
+ {250
RPS* rpsWriter = &curFrame->m_encData->m_slice->m_rps;251
int i, num = rpsWriter->numberOfPictures;252
char deltaPOC[128];253
x265_2.5.tar.gz/source/encoder/ratecontrol.h -> x265_2.6.tar.gz/source/encoder/ratecontrol.h
Changed
10
1
2
int rateControlStart(Frame* curFrame, RateControlEntry* rce, Encoder* enc);3
void rateControlUpdateStats(RateControlEntry* rce);4
int rateControlEnd(Frame* curFrame, int64_t bits, RateControlEntry* rce, int *filler);5
- int rowVbvRateControl(Frame* curFrame, uint32_t row, RateControlEntry* rce, double& qpVbv);6
+ int rowVbvRateControl(Frame* curFrame, uint32_t row, RateControlEntry* rce, double& qpVbv, uint32_t* m_sliceBaseRow, uint32_t sliceId);7
int rateControlSliceType(int frameNum);8
bool cuTreeReadFor2Pass(Frame* curFrame);9
void hrdFullness(SEIBufferingPeriod* sei);10
x265_2.5.tar.gz/source/encoder/search.cpp -> x265_2.6.tar.gz/source/encoder/search.cpp
Changed
10
1
2
3
/* Uni-directional prediction */4
if ((m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)5
- || (m_param->analysisMultiPassRefine && m_param->rc.bStatRead))6
+ || (m_param->analysisMultiPassRefine && m_param->rc.bStatRead) || (m_param->bMVType == AVC_INFO))7
{8
for (int list = 0; list < numPredDir; list++)9
{10
x265_2.5.tar.gz/source/encoder/slicetype.cpp -> x265_2.6.tar.gz/source/encoder/slicetype.cpp
Changed
255
1
2
m_filled = false;3
m_outputSignalRequired = false;4
m_isActive = true;5
+ m_inputCount = 0;6
7
m_8x8Height = ((m_param->sourceHeight / 2) + X265_LOWRES_CU_SIZE - 1) >> X265_LOWRES_CU_BITS;8
m_8x8Width = ((m_param->sourceWidth / 2) + X265_LOWRES_CU_SIZE - 1) >> X265_LOWRES_CU_BITS;9
10
/* Called by API thread */11
void Lookahead::addPicture(Frame& curFrame, int sliceType)12
{13
- curFrame.m_lowres.sliceType = sliceType;14
-15
- /* determine if the lookahead is (over) filled enough for frames to begin to16
- * be consumed by frame encoders */17
- if (!m_filled)18
+ if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->bDisableLookahead)19
{20
- if (!m_param->bframes & !m_param->lookaheadDepth)21
- m_filled = true; /* zero-latency */22
- else if (curFrame.m_poc >= m_param->lookaheadDepth + 2 + m_param->bframes)23
- m_filled = true; /* full capacity plus mini-gop lag */24
+ if (!m_filled)25
+ m_filled = true;26
+ m_outputLock.acquire();27
+ m_outputQueue.pushBack(curFrame);28
+ m_outputLock.release();29
+ m_inputCount++;30
+ }31
+ else32
+ {33
+ checkLookaheadQueue(m_inputCount);34
+ curFrame.m_lowres.sliceType = sliceType;35
+ addPicture(curFrame);36
}37
-38
- m_inputLock.acquire();39
- m_inputQueue.pushBack(curFrame);40
- if (m_pool && m_inputQueue.size() >= m_fullQueueSize)41
- tryWakeOne();42
- m_inputLock.release();43
}44
45
void Lookahead::addPicture(Frame& curFrame)46
47
m_inputLock.acquire();48
m_inputQueue.pushBack(curFrame);49
m_inputLock.release();50
+ m_inputCount++;51
}52
53
void Lookahead::checkLookaheadQueue(int &frameCnt)54
55
m_filled = true;56
}57
58
+void Lookahead::setLookaheadQueue()59
+{60
+ m_filled = false;61
+ m_fullQueueSize = X265_MAX(1, m_param->lookaheadDepth);62
+}63
+64
void Lookahead::findJob(int /*workerThreadID*/)65
{66
bool doDecide;67
68
m_outputLock.release();69
70
if (out)71
+ {72
+ m_inputCount--;73
return out;74
+ }75
+76
+ if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->bDisableLookahead)77
+ return NULL;78
79
findJob(-1); /* run slicetypeDecide() if necessary */80
81
82
if (wait)83
m_outputSignal.wait();84
85
- return m_outputQueue.popFront();86
+ out = m_outputQueue.popFront();87
+ if (out)88
+ m_inputCount--;89
+ return out;90
}91
else92
return NULL;93
94
default:95
return;96
}97
-98
- X265_CHECK(curFrame->m_lowres.costEst[b - p0][p1 - b] > 0, "Slice cost not estimated\n")99
-100
- if (m_param->rc.cuTree && !m_param->rc.bStatRead)101
- /* update row satds based on cutree offsets */102
- curFrame->m_lowres.satdCost = frameCostRecalculate(frames, p0, p1, b);103
- else if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || m_param->scaleFactor)104
- {105
- if (m_param->rc.aqMode)106
- curFrame->m_lowres.satdCost = curFrame->m_lowres.costEstAq[b - p0][p1 - b];107
- else108
- curFrame->m_lowres.satdCost = curFrame->m_lowres.costEst[b - p0][p1 - b];109
- }110
-111
- if (m_param->rc.vbvBufferSize && m_param->rc.vbvMaxBitrate)112
+ if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || !m_param->bDisableLookahead)113
{114
- /* aggregate lowres row satds to CTU resolution */115
- curFrame->m_lowres.lowresCostForRc = curFrame->m_lowres.lowresCosts[b - p0][p1 - b];116
- uint32_t lowresRow = 0, lowresCol = 0, lowresCuIdx = 0, sum = 0, intraSum = 0;117
- uint32_t scale = m_param->maxCUSize / (2 * X265_LOWRES_CU_SIZE);118
- uint32_t numCuInHeight = (m_param->sourceHeight + m_param->maxCUSize - 1) / m_param->maxCUSize;119
- uint32_t widthInLowresCu = (uint32_t)m_8x8Width, heightInLowresCu = (uint32_t)m_8x8Height;120
- double *qp_offset = 0;121
- /* Factor in qpoffsets based on Aq/Cutree in CU costs */122
- if (m_param->rc.aqMode || m_param->bAQMotion)123
- qp_offset = (frames[b]->sliceType == X265_TYPE_B || !m_param->rc.cuTree) ? frames[b]->qpAqOffset : frames[b]->qpCuTreeOffset;124
-125
- for (uint32_t row = 0; row < numCuInHeight; row++)126
+ X265_CHECK(curFrame->m_lowres.costEst[b - p0][p1 - b] > 0, "Slice cost not estimated\n")127
+ if (m_param->rc.cuTree && !m_param->rc.bStatRead)128
+ /* update row satds based on cutree offsets */129
+ curFrame->m_lowres.satdCost = frameCostRecalculate(frames, p0, p1, b);130
+ else if (m_param->analysisReuseMode != X265_ANALYSIS_LOAD || m_param->scaleFactor)131
+ {132
+ if (m_param->rc.aqMode)133
+ curFrame->m_lowres.satdCost = curFrame->m_lowres.costEstAq[b - p0][p1 - b];134
+ else135
+ curFrame->m_lowres.satdCost = curFrame->m_lowres.costEst[b - p0][p1 - b];136
+ }137
+ if (m_param->rc.vbvBufferSize && m_param->rc.vbvMaxBitrate)138
{139
- lowresRow = row * scale;140
- for (uint32_t cnt = 0; cnt < scale && lowresRow < heightInLowresCu; lowresRow++, cnt++)141
+ /* aggregate lowres row satds to CTU resolution */142
+ curFrame->m_lowres.lowresCostForRc = curFrame->m_lowres.lowresCosts[b - p0][p1 - b];143
+ uint32_t lowresRow = 0, lowresCol = 0, lowresCuIdx = 0, sum = 0, intraSum = 0;144
+ uint32_t scale = m_param->maxCUSize / (2 * X265_LOWRES_CU_SIZE);145
+ uint32_t numCuInHeight = (m_param->sourceHeight + m_param->maxCUSize - 1) / m_param->maxCUSize;146
+ uint32_t widthInLowresCu = (uint32_t)m_8x8Width, heightInLowresCu = (uint32_t)m_8x8Height;147
+ double *qp_offset = 0;148
+ /* Factor in qpoffsets based on Aq/Cutree in CU costs */149
+ if (m_param->rc.aqMode || m_param->bAQMotion)150
+ qp_offset = (frames[b]->sliceType == X265_TYPE_B || !m_param->rc.cuTree) ? frames[b]->qpAqOffset : frames[b]->qpCuTreeOffset;151
+152
+ for (uint32_t row = 0; row < numCuInHeight; row++)153
{154
- sum = 0; intraSum = 0;155
- int diff = 0;156
- lowresCuIdx = lowresRow * widthInLowresCu;157
- for (lowresCol = 0; lowresCol < widthInLowresCu; lowresCol++, lowresCuIdx++)158
+ lowresRow = row * scale;159
+ for (uint32_t cnt = 0; cnt < scale && lowresRow < heightInLowresCu; lowresRow++, cnt++)160
{161
- uint16_t lowresCuCost = curFrame->m_lowres.lowresCostForRc[lowresCuIdx] & LOWRES_COST_MASK;162
- if (qp_offset)163
+ sum = 0; intraSum = 0;164
+ int diff = 0;165
+ lowresCuIdx = lowresRow * widthInLowresCu;166
+ for (lowresCol = 0; lowresCol < widthInLowresCu; lowresCol++, lowresCuIdx++)167
{168
- double qpOffset;169
- if (m_param->rc.qgSize == 8)170
- qpOffset = (qp_offset[lowresCol * 2 + lowresRow * widthInLowresCu * 4] +171
- qp_offset[lowresCol * 2 + lowresRow * widthInLowresCu * 4 + 1] +172
- qp_offset[lowresCol * 2 + lowresRow * widthInLowresCu * 4 + curFrame->m_lowres.maxBlocksInRowFullRes] +173
- qp_offset[lowresCol * 2 + lowresRow * widthInLowresCu * 4 + curFrame->m_lowres.maxBlocksInRowFullRes + 1]) / 4;174
- else175
- qpOffset = qp_offset[lowresCuIdx];176
- lowresCuCost = (uint16_t)((lowresCuCost * x265_exp2fix8(qpOffset) + 128) >> 8);177
- int32_t intraCuCost = curFrame->m_lowres.intraCost[lowresCuIdx];178
- curFrame->m_lowres.intraCost[lowresCuIdx] = (intraCuCost * x265_exp2fix8(qpOffset) + 128) >> 8;179
+ uint16_t lowresCuCost = curFrame->m_lowres.lowresCostForRc[lowresCuIdx] & LOWRES_COST_MASK;180
+ if (qp_offset)181
+ {182
+ double qpOffset;183
+ if (m_param->rc.qgSize == 8)184
+ qpOffset = (qp_offset[lowresCol * 2 + lowresRow * widthInLowresCu * 4] +185
+ qp_offset[lowresCol * 2 + lowresRow * widthInLowresCu * 4 + 1] +186
+ qp_offset[lowresCol * 2 + lowresRow * widthInLowresCu * 4 + curFrame->m_lowres.maxBlocksInRowFullRes] +187
+ qp_offset[lowresCol * 2 + lowresRow * widthInLowresCu * 4 + curFrame->m_lowres.maxBlocksInRowFullRes + 1]) / 4;188
+ else189
+ qpOffset = qp_offset[lowresCuIdx];190
+ lowresCuCost = (uint16_t)((lowresCuCost * x265_exp2fix8(qpOffset) + 128) >> 8);191
+ int32_t intraCuCost = curFrame->m_lowres.intraCost[lowresCuIdx];192
+ curFrame->m_lowres.intraCost[lowresCuIdx] = (intraCuCost * x265_exp2fix8(qpOffset) + 128) >> 8;193
+ }194
+ if (m_param->bIntraRefresh && slice->m_sliceType == X265_TYPE_P)195
+ for (uint32_t x = curFrame->m_encData->m_pir.pirStartCol; x <= curFrame->m_encData->m_pir.pirEndCol; x++)196
+ diff += curFrame->m_lowres.intraCost[lowresCuIdx] - lowresCuCost;197
+ curFrame->m_lowres.lowresCostForRc[lowresCuIdx] = lowresCuCost;198
+ sum += lowresCuCost;199
+ intraSum += curFrame->m_lowres.intraCost[lowresCuIdx];200
}201
- if (m_param->bIntraRefresh && slice->m_sliceType == X265_TYPE_P)202
- for (uint32_t x = curFrame->m_encData->m_pir.pirStartCol; x <= curFrame->m_encData->m_pir.pirEndCol; x++)203
- diff += curFrame->m_lowres.intraCost[lowresCuIdx] - lowresCuCost;204
- curFrame->m_lowres.lowresCostForRc[lowresCuIdx] = lowresCuCost;205
- sum += lowresCuCost;206
- intraSum += curFrame->m_lowres.intraCost[lowresCuIdx];207
+ curFrame->m_encData->m_rowStat[row].satdForVbv += sum;208
+ curFrame->m_encData->m_rowStat[row].satdForVbv += diff;209
+ curFrame->m_encData->m_rowStat[row].intraSatdForVbv += intraSum;210
}211
- curFrame->m_encData->m_rowStat[row].satdForVbv += sum;212
- curFrame->m_encData->m_rowStat[row].satdForVbv += diff;213
- curFrame->m_encData->m_rowStat[row].intraSatdForVbv += intraSum;214
}215
}216
}217
218
(m_param->lookaheadDepth && m_param->rc.vbvBufferSize)))219
{220
slicetypeAnalyse(frames, false);221
+ bool bIsVbv = m_param->rc.vbvBufferSize > 0 && m_param->rc.vbvMaxBitrate > 0;222
+ if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->scaleFactor && bIsVbv)223
+ {224
+ int numFrames;225
+ for (numFrames = 0; numFrames < maxSearch; numFrames++)226
+ {227
+ Lowres *fenc = frames[numFrames + 1];228
+ if (!fenc)229
+ break;230
+ }231
+ vbvLookahead(frames, numFrames, true);232
+ }233
}234
235
int bframes, brefs;236
237
238
frames[j + 1] = NULL;239
slicetypeAnalyse(frames, true);240
+ bool bIsVbv = m_param->rc.vbvBufferSize > 0 && m_param->rc.vbvMaxBitrate > 0;241
+ if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->scaleFactor && bIsVbv)242
+ {243
+ int numFrames;244
+ for (numFrames = 0; numFrames < maxSearch; numFrames++)245
+ {246
+ Lowres *fenc = frames[numFrames + 1];247
+ if (!fenc)248
+ break;249
+ }250
+ vbvLookahead(frames, numFrames, true);251
+ }252
}253
m_outputLock.release();254
}255
x265_2.5.tar.gz/source/encoder/slicetype.h -> x265_2.6.tar.gz/source/encoder/slicetype.h
Changed
18
1
2
int m_cuCount;3
int m_numCoopSlices;4
int m_numRowsPerSlice;5
+ int m_inputCount;6
double m_cuTreeStrength;7
8
bool m_isActive;9
10
Frame* getDecidedPicture();11
12
void getEstimatedPictureCost(Frame *pic);13
-14
+ void setLookaheadQueue();15
16
protected:17
18
x265_2.5.tar.gz/source/input/y4m.cpp -> x265_2.6.tar.gz/source/input/y4m.cpp
Changed
41
1
2
break;3
}4
5
- switch (csp)6
+ if (csp / 100 == ('m'-'0')*1000 + ('o'-'0')*100 + ('n'-'0')*10 + ('o'-'0'))7
{8
- case ('m'-'0')*100000 + ('o'-'0')*10000 + ('n'-'0')*1000 + ('o'-'0')*100 + 16:9
colorSpace = X265_CSP_I400;10
- depth = 16;11
- break;12
-13
- case ('m'-'0')*1000 + ('o'-'0')*100 + ('n'-'0')*10 + ('o'-'0'):14
+ d = csp % 100;15
+ }16
+ else if (csp / 10 == ('m'-'0')*1000 + ('o'-'0')*100 + ('n'-'0')*10 + ('o'-'0'))17
+ {18
colorSpace = X265_CSP_I400;19
- depth = 8;20
- break;21
- 22
- default:23
- if (d >= 8 && d <= 16)24
- depth = d;25
- colorSpace = (csp == 444) ? X265_CSP_I444 : (csp == 422) ? X265_CSP_I422 : X265_CSP_I420;26
+ d = csp % 10;27
}28
+ else if (csp == ('m'-'0')*1000 + ('o'-'0')*100 + ('n'-'0')*10 + ('o'-'0'))29
+ {30
+ colorSpace = X265_CSP_I400;31
+ d = 8;32
+ }33
+ else34
+ colorSpace = (csp == 444) ? X265_CSP_I444 : (csp == 422) ? X265_CSP_I422 : X265_CSP_I420;35
+36
+ if (d >= 8 && d <= 16)37
+ depth = d;38
break;39
40
default:41
x265_2.5.tar.gz/source/test/rate-control-tests.txt -> x265_2.6.tar.gz/source/test/rate-control-tests.txt
Changed
57
1
2
3
#These tests should yeild deterministic results4
# This test is listed first since it currently reproduces bugs5
-big_buck_bunny_360p24.y4m,--preset medium --bitrate 1000 --pass 1 -F4,--preset medium --bitrate 1000 --pass 2 -F46
+big_buck_bunny_360p24.y4m,--preset medium --bitrate 1000 --pass 1 -F4::--preset medium --bitrate 1000 --pass 2 -F47
fire_1920x1080_30.yuv, --preset slow --bitrate 2000 --tune zero-latency 8
9
10
11
big_buck_bunny_360p24.y4m,--preset medium --bitrate 400 --vbv-bufsize 600 --vbv-maxrate 600 --no-wpp --aud --hrd --tune fast-decode12
sita_1920x1080_30.yuv,--preset superfast --bitrate 3000 --vbv-bufsize 3000 --vbv-maxrate 3000 --aud --strict-cbr --no-wpp13
sintel_trailer_2k_480p24.y4m, --preset slow --crf 24 --vbv-bufsize 150 --vbv-maxrate 150 --dynamic-rd 1.5314
-15
+BasketballDrive_1920x1080_50.y4m,--preset medium --bitrate 10000 --vbv-bufsize 15000 --vbv-maxrate 11500 --vbv-end 0.9 --vbv-end-fr-adj 0.716
17
18
# multi-pass rate control tests19
-sita_1920x1080_30.yuv, --preset ultrafast --crf 20 --no-cutree --no-scenecut --keyint 50 --no-open-gop --pass 1 --vbv-bufsize 7000 --vbv-maxrate 5000, --preset ultrafast --crf 20 --no-cutree --no-scenecut --keyint 50 --no-open-gop --pass 2 --vbv-bufsize 7000 --vbv-maxrate 500020
-sita_1920x1080_30.yuv, --preset medium --crf 20 --no-cutree --no-scenecut --keyint 50 --no-open-gop --pass 1 --vbv-bufsize 7000 --vbv-maxrate 5000, --preset medium --crf 20 --no-cutree --no-scenecut --keyint 50 --no-open-gop --pass 2 --vbv-bufsize 7000 --vbv-maxrate 500021
-sintel_trailer_2k_480p24.y4m, --preset medium --crf 18 --no-cutree --no-scenecut --no-open-gop --keyint 50 --vbv-bufsize 1200 --vbv-maxrate 1000 --pass 1, --preset medium --crf 18 --no-cutree --no-scenecut --no-open-gop --keyint 50 --vbv-bufsize 1200 --vbv-maxrate 1000 --pass 222
-sintel_trailer_2k_480p24.y4m, --preset veryslow --crf 18 --no-cutree --no-scenecut --no-open-gop --keyint 50 --vbv-bufsize 1200 --vbv-maxrate 1000 --pass 1, --preset veryslow --crf 18 --no-cutree --no-scenecut --no-open-gop --keyint 50 --vbv-bufsize 1200 --vbv-maxrate 1000 --pass 223
-ten_teaser_3840x2160_50_10bit.yuv, --preset medium --crf 25 --no-cutree --no-open-gop --no-scenecut --keyint 50 --vbv-maxrate 10000 --vbv-bufsize 12000 --pass 1, --preset medium --crf 25 --no-cutree --no-open-gop --no-scenecut --keyint 50 --vbv-maxrate 10000 --vbv-bufsize 12000 --pass 224
-big_buck_bunny_360p24.y4m,--preset slow --crf 40 --pass 1 -f 5000,--preset slow --bitrate 200 --pass 2 -f 500025
-big_buck_bunny_360p24.y4m,--preset medium --bitrate 700 --pass 1 -F4 --slow-firstpass -f 5000 ,--preset medium --bitrate 700 --vbv-bufsize 900 --vbv-maxrate 700 --pass 2 -F4 -f 500026
-112_1920x1080_25.yuv,--preset fast --bitrate 1000 --vbv-maxrate 1000 --vbv-bufsize 1000 --strict-cbr --pass 1 -F4,--preset fast --bitrate 1000 --vbv-maxrate 3000 --vbv-bufsize 3000 --pass 2 -F427
-pine_tree_1920x1080_30.yuv,--preset veryfast --crf 12 --pass 1 -F4,--preset faster --bitrate 4000 --pass 2 -F428
-SteamLocomotiveTrain_2560x1600_60_10bit_crop.yuv, --tune grain --preset ultrafast --bitrate 5000 --vbv-maxrate 5000 --vbv-bufsize 8000 --strict-cbr -F4 --pass 1, --tune grain --preset ultrafast --bitrate 8000 --vbv-maxrate 8000 --vbv-bufsize 8000 -F4 --pass 229
-RaceHorses_416x240_30_10bit.yuv,--preset medium --crf 40 --pass 1, --preset faster --bitrate 200 --pass 2 -F430
-CrowdRun_1920x1080_50_10bit_422.yuv,--preset superfast --bitrate 2500 --pass 1 -F4 --slow-firstpass,--preset superfast --bitrate 2500 --pass 2 -F431
-RaceHorses_416x240_30_10bit.yuv,--preset medium --crf 26 --vbv-maxrate 1000 --vbv-bufsize 1000 --pass 1,--preset fast --bitrate 1000 --vbv-maxrate 1000 --vbv-bufsize 700 --pass 3 -F4,--preset slow --bitrate 500 --vbv-maxrate 500 --vbv-bufsize 700 --pass 2 -F432
-sita_1920x1080_30.yuv, --preset ultrafast --crf 20 --no-cutree --keyint 50 --min-keyint 50 --no-open-gop --pass 1 --vbv-bufsize 7000 --vbv-maxrate 5000, --preset ultrafast --crf 20 --no-cutree --keyint 50 --min-keyint 50 --no-open-gop --pass 2 --vbv-bufsize 7000 --vbv-maxrate 5000 --repeat-headers33
-sita_1920x1080_30.yuv, --preset medium --crf 20 --no-cutree --keyint 50 --min-keyint 50 --no-open-gop --pass 1 --vbv-bufsize 7000 --vbv-maxrate 5000 --repeat-headers --multi-pass-opt-rps, --preset medium --crf 20 --no-cutree --keyint 50 --min-keyint 50 --no-open-gop --pass 2 --vbv-bufsize 7000 --vbv-maxrate 5000 --repeat-headers --multi-pass-opt-rps34
+sita_1920x1080_30.yuv, --preset ultrafast --crf 20 --no-cutree --no-scenecut --keyint 50 --no-open-gop --pass 1 --vbv-bufsize 7000 --vbv-maxrate 5000:: --preset ultrafast --crf 20 --no-cutree --no-scenecut --keyint 50 --no-open-gop --pass 2 --vbv-bufsize 7000 --vbv-maxrate 500035
+sita_1920x1080_30.yuv, --preset medium --crf 20 --no-cutree --no-scenecut --keyint 50 --no-open-gop --pass 1 --vbv-bufsize 7000 --vbv-maxrate 5000:: --preset medium --crf 20 --no-cutree --no-scenecut --keyint 50 --no-open-gop --pass 2 --vbv-bufsize 7000 --vbv-maxrate 500036
+sintel_trailer_2k_480p24.y4m, --preset medium --crf 18 --no-cutree --no-scenecut --no-open-gop --keyint 50 --vbv-bufsize 1200 --vbv-maxrate 1000 --pass 1:: --preset medium --crf 18 --no-cutree --no-scenecut --no-open-gop --keyint 50 --vbv-bufsize 1200 --vbv-maxrate 1000 --pass 237
+sintel_trailer_2k_480p24.y4m, --preset veryslow --crf 18 --no-cutree --no-scenecut --no-open-gop --keyint 50 --vbv-bufsize 1200 --vbv-maxrate 1000 --pass 1:: --preset veryslow --crf 18 --no-cutree --no-scenecut --no-open-gop --keyint 50 --vbv-bufsize 1200 --vbv-maxrate 1000 --pass 238
+ten_teaser_3840x2160_50_10bit.yuv, --preset medium --crf 25 --no-cutree --no-open-gop --no-scenecut --keyint 50 --vbv-maxrate 10000 --vbv-bufsize 12000 --pass 1:: --preset medium --crf 25 --no-cutree --no-open-gop --no-scenecut --keyint 50 --vbv-maxrate 10000 --vbv-bufsize 12000 --pass 239
+big_buck_bunny_360p24.y4m,--preset slow --crf 40 --pass 1 -f 5000::--preset slow --bitrate 200 --pass 2 -f 500040
+big_buck_bunny_360p24.y4m,--preset medium --bitrate 700 --pass 1 -F4 --slow-firstpass -f 5000 ::--preset medium --bitrate 700 --vbv-bufsize 900 --vbv-maxrate 700 --pass 2 -F4 -f 500041
+112_1920x1080_25.yuv,--preset fast --bitrate 1000 --vbv-maxrate 1000 --vbv-bufsize 1000 --strict-cbr --pass 1 -F4::--preset fast --bitrate 1000 --vbv-maxrate 3000 --vbv-bufsize 3000 --pass 2 -F442
+pine_tree_1920x1080_30.yuv,--preset veryfast --crf 12 --pass 1 -F4::--preset faster --bitrate 4000 --pass 2 -F443
+SteamLocomotiveTrain_2560x1600_60_10bit_crop.yuv, --tune grain --preset ultrafast --bitrate 5000 --vbv-maxrate 5000 --vbv-bufsize 8000 --strict-cbr -F4 --pass 1:: --tune grain --preset ultrafast --bitrate 8000 --vbv-maxrate 8000 --vbv-bufsize 8000 -F4 --pass 244
+RaceHorses_416x240_30_10bit.yuv,--preset medium --crf 40 --pass 1:: --preset faster --bitrate 200 --pass 2 -F445
+CrowdRun_1920x1080_50_10bit_422.yuv,--preset superfast --bitrate 2500 --pass 1 -F4 --slow-firstpass::--preset superfast --bitrate 2500 --pass 2 -F446
+RaceHorses_416x240_30_10bit.yuv,--preset medium --crf 26 --vbv-maxrate 1000 --vbv-bufsize 1000 --pass 1::--preset fast --bitrate 1000 --vbv-maxrate 1000 --vbv-bufsize 700 --pass 3 -F4::--preset slow --bitrate 500 --vbv-maxrate 500 --vbv-bufsize 700 --pass 2 -F447
+sita_1920x1080_30.yuv, --preset ultrafast --crf 20 --no-cutree --keyint 50 --min-keyint 50 --no-open-gop --pass 1 --vbv-bufsize 7000 --vbv-maxrate 5000:: --preset ultrafast --crf 20 --no-cutree --keyint 50 --min-keyint 50 --no-open-gop --pass 2 --vbv-bufsize 7000 --vbv-maxrate 5000 --repeat-headers48
+sita_1920x1080_30.yuv, --preset medium --crf 20 --no-cutree --keyint 50 --min-keyint 50 --no-open-gop --pass 1 --vbv-bufsize 7000 --vbv-maxrate 5000 --repeat-headers --multi-pass-opt-rps:: --preset medium --crf 20 --no-cutree --keyint 50 --min-keyint 50 --no-open-gop --pass 2 --vbv-bufsize 7000 --vbv-maxrate 5000 --repeat-headers --multi-pass-opt-rps49
50
# multi-pass rate control and analysis51
-ducks_take_off_1080p50.y4m,--bitrate 6000 --pass 1 --multi-pass-opt-analysis --hash 1 --ssim --psnr, --bitrate 6000 --pass 2 --multi-pass-opt-analysis --hash 1 --ssim --psnr52
-big_buck_bunny_360p24.y4m,--preset veryslow --bitrate 600 --pass 1 --multi-pass-opt-analysis --multi-pass-opt-distortion --hash 1 --ssim --psnr, --preset veryslow --bitrate 600 --pass 2 --multi-pass-opt-analysis --multi-pass-opt-distortion --hash 1 --ssim --psnr53
-parkrun_ter_720p50.y4m, --bitrate 3500 --pass 1 --multi-pass-opt-distortion --hash 1 --ssim --psnr, --bitrate 3500 --pass 3 --multi-pass-opt-distortion --hash 1 --ssim --psnr, --bitrate 3500 --pass 2 --multi-pass-opt-distortion --hash 1 --ssim --psnr54
+ducks_take_off_1080p50.y4m,--bitrate 6000 --pass 1 --multi-pass-opt-analysis --hash 1 --ssim --psnr:: --bitrate 6000 --pass 2 --multi-pass-opt-analysis --hash 1 --ssim --psnr55
+big_buck_bunny_360p24.y4m,--preset veryslow --bitrate 600 --pass 1 --multi-pass-opt-analysis --multi-pass-opt-distortion --hash 1 --ssim --psnr:: --preset veryslow --bitrate 600 --pass 2 --multi-pass-opt-analysis --multi-pass-opt-distortion --hash 1 --ssim --psnr56
+parkrun_ter_720p50.y4m, --bitrate 3500 --pass 1 --multi-pass-opt-distortion --hash 1 --ssim --psnr:: --bitrate 3500 --pass 3 --multi-pass-opt-distortion --hash 1 --ssim --psnr:: --bitrate 3500 --pass 2 --multi-pass-opt-distortion --hash 1 --ssim --psnr57
x265_2.5.tar.gz/source/test/regression-tests.txt -> x265_2.6.tar.gz/source/test/regression-tests.txt
Changed
74
1
2
3
BasketballDrive_1920x1080_50.y4m,--preset ultrafast --signhide --colormatrix bt7094
BasketballDrive_1920x1080_50.y4m,--preset superfast --psy-rd 1 --ctu 16 --no-wpp --limit-modes5
+BasketballDrive_1920x1080_50.y4m,--preset superfast --tune zerolatency --bitrate 9000 --vbv-maxrate 9000 --vbv-bufsize 9000 -F 1 --slices 26
BasketballDrive_1920x1080_50.y4m,--preset veryfast --tune zerolatency --no-temporal-mvp7
BasketballDrive_1920x1080_50.y4m,--preset faster --aq-strength 2 --merange 190 --slices 38
BasketballDrive_1920x1080_50.y4m,--preset medium --ctu 16 --max-tu-size 8 --subme 7 --qg-size 16 --cu-lossless --tu-inter-depth 3 --limit-tu 19
BasketballDrive_1920x1080_50.y4m,--preset medium --keyint -1 --nr-inter 100 -F4 --no-sao10
-BasketballDrive_1920x1080_50.y4m,--preset medium --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 2 --bitrate 7000 --limit-modes,--preset medium --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 2 --bitrate 7000 --limit-modes11
+BasketballDrive_1920x1080_50.y4m,--preset medium --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 2 --bitrate 7000 --limit-modes::--preset medium --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 2 --bitrate 7000 --limit-modes12
BasketballDrive_1920x1080_50.y4m,--preset slow --nr-intra 100 -F4 --aq-strength 3 --qg-size 16 --limit-refs 113
BasketballDrive_1920x1080_50.y4m,--preset slower --lossless --chromaloc 3 --subme 0 --limit-tu 414
-BasketballDrive_1920x1080_50.y4m,--preset slower --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 10 --bitrate 7000 --limit-tu 0,--preset slower --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 10 --bitrate 7000 --limit-tu 015
+BasketballDrive_1920x1080_50.y4m,--preset slower --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 10 --bitrate 7000 --limit-tu 0::--preset slower --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 10 --bitrate 7000 --limit-tu 016
BasketballDrive_1920x1080_50.y4m,--preset veryslow --crf 4 --cu-lossless --pmode --limit-refs 1 --aq-mode 3 --limit-tu 317
-BasketballDrive_1920x1080_50.y4m,--preset veryslow --no-cutree --analysis-reuse-mode=save --bitrate 7000 --tskip-fast --limit-tu 4,--preset veryslow --no-cutree --analysis-reuse-mode=load --bitrate 7000 --tskip-fast --limit-tu 418
+BasketballDrive_1920x1080_50.y4m,--preset veryslow --no-cutree --analysis-reuse-mode=save --bitrate 7000 --tskip-fast --limit-tu 4::--preset veryslow --no-cutree --analysis-reuse-mode=load --bitrate 7000 --tskip-fast --limit-tu 419
BasketballDrive_1920x1080_50.y4m,--preset veryslow --recon-y4m-exec "ffplay -i pipe:0 -autoexit"20
Coastguard-4k.y4m,--preset ultrafast --recon-y4m-exec "ffplay -i pipe:0 -autoexit"21
Coastguard-4k.y4m,--preset superfast --tune grain --overscan=crop22
Coastguard-4k.y4m,--preset superfast --tune grain --pme --aq-strength 2 --merange 19023
-Coastguard-4k.y4m,--preset veryfast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 1 --bitrate 15000,--preset veryfast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 1 --bitrate 1500024
+Coastguard-4k.y4m,--preset veryfast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 1 --bitrate 15000::--preset veryfast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 1 --bitrate 1500025
Coastguard-4k.y4m,--preset medium --rdoq-level 1 --tune ssim --no-signhide --me umh --slices 226
Coastguard-4k.y4m,--preset slow --tune psnr --cbqpoffs -1 --crqpoffs 1 --limit-refs 127
CrowdRun_1920x1080_50_10bit_422.yuv,--preset ultrafast --weightp --tune zerolatency --qg-size 1628
29
DucksAndLegs_1920x1080_60_10bit_444.yuv,--preset veryfast --weightp --nr-intra 1000 -F430
DucksAndLegs_1920x1080_60_10bit_444.yuv,--preset medium --nr-inter 500 -F4 --no-psy-rdoq31
DucksAndLegs_1920x1080_60_10bit_444.yuv,--preset slower --no-weightp --rdoq-level 0 --limit-refs 3 --tu-inter-depth 4 --limit-tu 332
-DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset fast --no-cutree --analysis-reuse-mode=save --bitrate 3000 --early-skip --tu-inter-depth 3 --limit-tu 1,--preset fast --no-cutree --analysis-reuse-mode=load --bitrate 3000 --early-skip --tu-inter-depth 3 --limit-tu 133
+DucksAndLegs_1920x1080_60_10bit_422.yuv,--preset fast --no-cutree --analysis-reuse-mode=save --bitrate 3000 --early-skip --tu-inter-depth 3 --limit-tu 1::--preset fast --no-cutree --analysis-reuse-mode=load --bitrate 3000 --early-skip --tu-inter-depth 3 --limit-tu 134
FourPeople_1280x720_60.y4m,--preset superfast --no-wpp --lookahead-slices 235
FourPeople_1280x720_60.y4m,--preset veryfast --aq-mode 2 --aq-strength 1.5 --qg-size 836
FourPeople_1280x720_60.y4m,--preset medium --qp 38 --no-psy-rd37
38
KristenAndSara_1280x720_60.y4m,--preset slower --pmode --max-tu-size 8 --limit-refs 0 --limit-modes --limit-tu 139
NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset superfast --tune psnr40
NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset medium --tune grain --limit-refs 241
-NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset slow --no-cutree --analysis-reuse-mode=save --rd 5 --analysis-reuse-level 10 --bitrate 9000,--preset slow --no-cutree --analysis-reuse-mode=load --rd 5 --analysis-reuse-level 10 --bitrate 900042
-News-4k.y4m,--preset ultrafast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 2 --bitrate 15000,--preset ultrafast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 2 --bitrate 1500043
+NebutaFestival_2560x1600_60_10bit_crop.yuv,--preset slow --no-cutree --analysis-reuse-mode=save --rd 5 --analysis-reuse-level 10 --bitrate 9000::--preset slow --no-cutree --analysis-reuse-mode=load --rd 5 --analysis-reuse-level 10 --bitrate 900044
+News-4k.y4m,--preset ultrafast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 2 --bitrate 15000::--preset ultrafast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 2 --bitrate 1500045
News-4k.y4m,--preset superfast --lookahead-slices 6 --aq-mode 046
News-4k.y4m,--preset superfast --slices 4 --aq-mode 0 47
News-4k.y4m,--preset medium --tune ssim --no-sao --qg-size 1648
49
ducks_take_off_444_720p50.y4m,--preset superfast --weightp --limit-refs 250
ducks_take_off_420_720p50.y4m,--preset faster --qp 24 --deblock -6 --limit-refs 251
ducks_take_off_420_720p50.y4m,--preset fast --deblock 6 --bframes 16 --rc-lookahead 4052
+ducks_take_off_420_720p50.y4m,--preset fast --tune zerolatency --crf 21 --vbv-maxrate 6000 --vbv-bufsize 6000 -F 1 --slices 253
ducks_take_off_420_720p50.y4m,--preset medium --tskip --tskip-fast --constrained-intra54
ducks_take_off_444_720p50.y4m,--preset medium --qp 38 --no-scenecut55
ducks_take_off_420_720p50.y4m,--preset slow --scaling-list default --qp 4056
57
old_town_cross_444_720p50.y4m,--preset superfast --weightp --min-cu 16 --limit-modes58
old_town_cross_444_720p50.y4m,--preset veryfast --qp 1 --tune ssim59
old_town_cross_444_720p50.y4m,--preset faster --rd 1 --tune zero-latency60
-old_town_cross_444_720p50.y4m,--preset fast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 1 --bitrate 3000 --early-skip,--preset fast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 1 --bitrate 3000 --early-skip61
+old_town_cross_444_720p50.y4m,--preset fast --no-cutree --analysis-reuse-mode=save --analysis-reuse-level 1 --bitrate 3000 --early-skip::--preset fast --no-cutree --analysis-reuse-mode=load --analysis-reuse-level 1 --bitrate 3000 --early-skip62
old_town_cross_444_720p50.y4m,--preset medium --keyint -1 --no-weightp --ref 663
old_town_cross_444_720p50.y4m,--preset slow --rdoq-level 1 --early-skip --ref 7 --no-b-pyramid64
old_town_cross_444_720p50.y4m,--preset slower --crf 4 --cu-lossless65
66
#SEA Implementation Test67
silent_cif_420.y4m,--preset veryslow --me sea68
big_buck_bunny_360p24.y4m,--preset superfast --me sea69
+70
+#low-pass dct test71
+720p50_parkrun_ter.y4m,--preset medium --lowpass-dct72
+73
# vim: tw=20074
x265_2.5.tar.gz/source/test/testharness.h -> x265_2.6.tar.gz/source/test/testharness.h
Changed
12
1
2
#include <intrin.h>3
#elif HAVE_RDTSC4
#include <intrin.h>5
+#elif (!defined(__APPLE__) && (defined (__GNUC__) && (defined(__x86_64__) || defined(__i386__))))6
+#include <x86intrin.h>7
+#elif ( !defined(__APPLE__) && defined (__GNUC__) && defined(__ARM_NEON__))8
+#include <arm_neon.h>9
#elif defined(__GNUC__)10
/* fallback for older GCC/MinGW */11
static inline uint32_t __rdtsc(void)12
x265_2.5.tar.gz/source/x265.cpp -> x265_2.6.tar.gz/source/x265.cpp
Changed
18
1
2
#endif3
4
#include "x265.h"5
-#include "x265-extras.h"6
#include "x265cli.h"7
8
#include "input/input.h"9
10
{11
if (pic_in->bitDepth > param->internalBitDepth && cliopt.bDither)12
{13
- x265_dither_image(*api, *pic_in, cliopt.input->getWidth(), cliopt.input->getHeight(), errorBuf, param->internalBitDepth);14
+ x265_dither_image(pic_in, cliopt.input->getWidth(), cliopt.input->getHeight(), errorBuf, param->internalBitDepth);15
pic_in->bitDepth = param->internalBitDepth;16
}17
/* Overwrite PTS */18
x265_2.5.tar.gz/source/x265.def.in -> x265_2.6.tar.gz/source/x265.def.in
Changed
13
1
2
x265_api_get_${X265_BUILD}3
x265_api_query4
x265_encoder_intra_refresh5
+x265_encoder_ctu_info6
+x265_get_slicetype_poc_and_scenecut7
+x265_get_ref_frame_list8
+x265_csvlog_open9
+x265_csvlog_frame10
+x265_csvlog_encode11
+x265_dither_image12
+x265_set_analysis_data13
x265_2.5.tar.gz/source/x265.h -> x265_2.6.tar.gz/source/x265.h
Changed
177
1
2
* opaque handler for encoder */3
typedef struct x265_encoder x265_encoder;4
5
+/* x265_picyuv:6
+ * opaque handler for PicYuv */7
+typedef struct x265_picyuv x265_picyuv;8
+9
/* Application developers planning to link against a shared library version of10
* libx265 from a Microsoft Visual Studio or similar development environment11
* will need to define X265_API_IMPORTS before including this header.12
13
uint8_t* payload;14
} x265_nal;15
16
+#define X265_LOOKAHEAD_MAX 25017
+18
+typedef struct x265_lookahead_data19
+{20
+ int64_t plannedSatd[X265_LOOKAHEAD_MAX + 1];21
+ uint32_t *vbvCost;22
+ uint32_t *intraVbvCost;23
+ uint32_t *satdForVbv;24
+ uint32_t *intraSatdForVbv;25
+ int keyframe;26
+ int lastMiniGopBFrame;27
+ int plannedType[X265_LOOKAHEAD_MAX + 1];28
+ int64_t dts;29
+} x265_lookahead_data;30
+31
/* Stores all analysis data for a single frame */32
typedef struct x265_analysis_data33
{34
35
void* wt;36
void* interData;37
void* intraData;38
+ uint32_t numCuInHeight;39
+ x265_lookahead_data lookahead;40
+ uint8_t* modeFlag[2];41
} x265_analysis_data;42
43
/* cu statistics */44
45
CTU_INFO_CHANGE = 2,46
}CTUInfo;47
48
+typedef enum49
+{50
+ NO_INFO = 0,51
+ AVC_INFO = 1,52
+}MVRefineType;53
54
/* Arbitrary User SEI55
* Payload size is in bytes and the payload pointer must be non-NULL. 56
57
/* String values accepted by x265_param_parse() (and CLI) for various parameters */58
static const char * const x265_motion_est_names[] = { "dia", "hex", "umh", "star", "sea", "full", 0 };59
static const char * const x265_source_csp_names[] = { "i400", "i420", "i422", "i444", "nv12", "nv16", 0 };60
-static const char * const x265_video_format_names[] = { "component", "pal", "ntsc", "secam", "mac", "undef", 0 };61
+static const char * const x265_video_format_names[] = { "component", "pal", "ntsc", "secam", "mac", "unknown", 0 };62
static const char * const x265_fullrange_names[] = { "limited", "full", 0 };63
-static const char * const x265_colorprim_names[] = { "", "bt709", "undef", "", "bt470m", "bt470bg", "smpte170m", "smpte240m", "film", "bt2020", 0 };64
-static const char * const x265_transfer_names[] = { "", "bt709", "undef", "", "bt470m", "bt470bg", "smpte170m", "smpte240m", "linear", "log100",65
+static const char * const x265_colorprim_names[] = { "reserved", "bt709", "unknown", "reserved", "bt470m", "bt470bg", "smpte170m", "smpte240m", "film", "bt2020", "smpte428", "smpte431", "smpte432", 0 };66
+static const char * const x265_transfer_names[] = { "reserved", "bt709", "unknown", "reserved", "bt470m", "bt470bg", "smpte170m", "smpte240m", "linear", "log100",67
"log316", "iec61966-2-4", "bt1361e", "iec61966-2-1", "bt2020-10", "bt2020-12",68
- "smpte-st-2084", "smpte-st-428", "arib-std-b67", 0 };69
-static const char * const x265_colmatrix_names[] = { "GBR", "bt709", "undef", "", "fcc", "bt470bg", "smpte170m", "smpte240m",70
- "YCgCo", "bt2020nc", "bt2020c", 0 };71
-static const char * const x265_sar_names[] = { "undef", "1:1", "12:11", "10:11", "16:11", "40:33", "24:11", "20:11",72
+ "smpte2084", "smpte428", "arib-std-b67", 0 };73
+static const char * const x265_colmatrix_names[] = { "gbr", "bt709", "unknown", "", "fcc", "bt470bg", "smpte170m", "smpte240m",74
+ "ycgco", "bt2020nc", "bt2020c", "smpte2085", "chroma-derived-nc", "chroma-derived-c", "ictcp", 0 };75
+static const char * const x265_sar_names[] = { "unknown", "1:1", "12:11", "10:11", "16:11", "40:33", "24:11", "20:11",76
"32:11", "80:33", "18:11", "15:11", "64:33", "160:99", "4:3", "3:2", "2:1", 0 };77
static const char * const x265_interlace_names[] = { "prog", "tff", "bff", 0 };78
static const char * const x265_analysis_names[] = { "off", "save", "load", 0 };79
80
81
/* File pointer for csv log */82
FILE* csvfpt;83
+84
+ /* Force flushing the frames from encoder */85
+ int forceFlush;86
+87
+ /* Enable skipping split RD analysis when sum of split CU rdCost larger than none split CU rdCost for Intra CU */88
+ int bEnableSplitRdSkip;89
+90
+ /* Disable lookahead */91
+ int bDisableLookahead;92
+93
+ /* Use low-pass subband dct approximation 94
+ * This DCT approximation is less computational intensive and gives results close to standard DCT */95
+ int bLowPassDct;96
+97
+ /* Sets the portion of the decode buffer that must be available after all the98
+ * specified frames have been inserted into the decode buffer. If it is less99
+ * than 1, then the final buffer available is vbv-end * vbvBufferSize. Otherwise,100
+ * it is interpreted as the final buffer available in kbits. Default 0 (disabled) */101
+ double vbvBufferEnd;102
+ 103
+ /* Frame from which qp has to be adjusted to hit final decode buffer emptiness.104
+ * Specified as a fraction of the total frames. Default 0 */105
+ double vbvEndFrameAdjust;106
+107
+ /* Reuse MV information obtained through API */108
+ int bMVType;109
+110
+ /* Allow the encoder to have a copy of the planes of x265_picture in Frame */111
+ int bCopyPicToFrame;112
} x265_param;113
114
/* x265_param_alloc:115
116
* the encoder will wait for this copy to complete if enabled.117
*/118
int x265_encoder_ctu_info(x265_encoder *, int poc, x265_ctu_info_t** ctu);119
+120
+/* x265_get_slicetype_poc_and_scenecut:121
+ * get the slice type, poc and scene cut information for the current frame,122
+ * returns negative on error, 0 when access unit were output.123
+ * This API must be called after(poc >= lookaheadDepth + bframes + 2) condition check */124
+int x265_get_slicetype_poc_and_scenecut(x265_encoder *encoder, int *slicetype, int *poc, int* sceneCut);125
+126
+/* x265_get_ref_frame_list:127
+ * returns negative on error, 0 when access unit were output.128
+ * This API must be called after(poc >= lookaheadDepth + bframes + 2) condition check */129
+int x265_get_ref_frame_list(x265_encoder *encoder, x265_picyuv**, x265_picyuv**, int, int);130
+131
+/* x265_set_analysis_data:132
+ * set the analysis data. The incoming analysis_data structure is assumed to be AVC-sized blocks.133
+ * returns negative on error, 0 access unit were output. */134
+int x265_set_analysis_data(x265_encoder *encoder, x265_analysis_data *analysis_data, int poc, uint32_t cuBytes);135
+136
/* x265_cleanup:137
* release library static allocations, reset configured CTU size */138
void x265_cleanup(void);139
140
+/* Open a CSV log file. On success it returns a file handle which must be passed141
+ * to x265_csvlog_frame() and/or x265_csvlog_encode(). The file handle must be142
+ * closed by the caller using fclose(). If csv-loglevel is 0, then no frame logging143
+ * header is written to the file. This function will return NULL if it is unable144
+ * to open the file for write or if it detects a structure size skew */145
+FILE* x265_csvlog_open(const x265_param *);146
+147
+/* Log frame statistics to the CSV file handle. csv-loglevel should have been non-zero148
+ * in the call to x265_csvlog_open() if this function is called. */149
+void x265_csvlog_frame(const x265_param *, const x265_picture *);150
+151
+/* Log final encode statistics to the CSV file handle. 'argc' and 'argv' are152
+ * intended to be command line arguments passed to the encoder. Encode153
+ * statistics should be queried from the encoder just prior to closing it. */154
+void x265_csvlog_encode(x265_encoder *encoder, const x265_stats *, int argc, char** argv);155
+156
+/* In-place downshift from a bit-depth greater than 8 to a bit-depth of 8, using157
+ * the residual bits to dither each row. */158
+void x265_dither_image(x265_picture *, int picWidth, int picHeight, int16_t *errorBuf, int bitDepth);159
+160
#define X265_MAJOR_VERSION 1161
162
/* === Multi-lib API ===163
164
int sizeof_frame_stats; /* sizeof(x265_frame_stats) */165
int (*encoder_intra_refresh)(x265_encoder*);166
int (*encoder_ctu_info)(x265_encoder*, int, x265_ctu_info_t**);167
+ int (*get_slicetype_poc_and_scenecut)(x265_encoder*, int*, int*, int*);168
+ int (*get_ref_frame_list)(x265_encoder*, x265_picyuv**, x265_picyuv**, int, int);169
+ FILE* (*csvlog_open)(const x265_param*);170
+ void (*csvlog_frame)(const x265_param*, const x265_picture*);171
+ void (*csvlog_encode)(x265_encoder*, const x265_stats*, int, char**);172
+ void (*dither_image)(x265_picture*, int, int, int16_t*, int);173
+ int (*set_analysis_data)(x265_encoder *encoder, x265_analysis_data *analysis_data, int poc, uint32_t cuBytes);174
/* add new pointers to the end, or increment X265_MAJOR_VERSION */175
} x265_api;176
177
x265_2.5.tar.gz/source/x265cli.h -> x265_2.6.tar.gz/source/x265cli.h
Changed
130
1
2
{ "vbv-maxrate", required_argument, NULL, 0 },3
{ "vbv-bufsize", required_argument, NULL, 0 },4
{ "vbv-init", required_argument, NULL, 0 },5
+ { "vbv-end", required_argument, NULL, 0 },6
+ { "vbv-end-fr-adj", required_argument, NULL, 0 },7
{ "bitrate", required_argument, NULL, 0 },8
{ "qp", required_argument, NULL, 'q' },9
{ "aq-mode", required_argument, NULL, 0 },10
11
{ "analysis-reuse-level", required_argument, NULL, 0 },12
{ "scale-factor", required_argument, NULL, 0 },13
{ "refine-intra", required_argument, NULL, 0 },14
- { "refine-inter", no_argument, NULL, 0 },15
- { "no-refine-inter",no_argument, NULL, 0 },16
+ { "refine-inter", required_argument, NULL, 0 },17
{ "strict-cbr", no_argument, NULL, 0 },18
{ "temporal-layers", no_argument, NULL, 0 },19
{ "no-temporal-layers", no_argument, NULL, 0 },20
21
{ "no-dhdr10-opt", no_argument, NULL, 0},22
{ "refine-mv", no_argument, NULL, 0 },23
{ "no-refine-mv", no_argument, NULL, 0 },24
+ { "force-flush", required_argument, NULL, 0 },25
+ { "splitrd-skip", no_argument, NULL, 0 },26
+ { "no-splitrd-skip", no_argument, NULL, 0 },27
+ { "lowpass-dct", no_argument, NULL, 0 },28
+ { "refine-mv-type", required_argument, NULL, 0 },29
+ { "copy-pic", no_argument, NULL, 0 },30
+ { "no-copy-pic", no_argument, NULL, 0 },31
{ 0, 0, 0, 0 },32
{ 0, 0, 0, 0 },33
{ 0, 0, 0, 0 },34
35
H0(" --seek <integer> First frame to encode\n");36
H1(" --[no-]interlace <bff|tff> Indicate input pictures are interlace fields in temporal order. Default progressive\n");37
H1(" --dither Enable dither if downscaling to 8 bit pixels. Default disabled\n");38
+ H0(" --[no-]copy-pic Copy buffers of input picture in frame. Default %s\n", OPT(param->bCopyPicToFrame));39
H0("\nQuality reporting metrics:\n");40
H0(" --[no-]ssim Enable reporting SSIM metric scores. Default %s\n", OPT(param->bEnableSsim));41
H0(" --[no-]psnr Enable reporting PSNR metric scores. Default %s\n", OPT(param->bEnablePsnr));42
43
H0(" --[no-]early-skip Enable early SKIP detection. Default %s\n", OPT(param->bEnableEarlySkip));44
H0(" --[no-]rskip Enable early exit from recursion. Default %s\n", OPT(param->bEnableRecursionSkip));45
H1(" --[no-]tskip-fast Enable fast intra transform skipping. Default %s\n", OPT(param->bEnableTSkipFast));46
+ H1(" --[no-]splitrd-skip Enable skipping split RD analysis when sum of split CU rdCost larger than none split CU rdCost for Intra CU. Default %s\n", OPT(param->bEnableSplitRdSkip));47
H1(" --nr-intra <integer> An integer value in range of 0 to 2000, which denotes strength of noise reduction in intra CUs. Default 0\n");48
H1(" --nr-inter <integer> An integer value in range of 0 to 2000, which denotes strength of noise reduction in inter CUs. Default 0\n");49
H0(" --ctu-info <integer> Enable receiving ctu information asynchronously and determine reaction to the CTU information (0, 1, 2, 4, 6) Default 0\n"50
51
H0(" --rc-lookahead <integer> Number of frames for frame-type lookahead (determines encoder latency) Default %d\n", param->lookaheadDepth);52
H1(" --lookahead-slices <0..16> Number of slices to use per lookahead cost estimate. Default %d\n", param->lookaheadSlices);53
H0(" --lookahead-threads <integer> Number of threads to be dedicated to perform lookahead only. Default %d\n", param->lookaheadThreads);54
- H0(" --bframes <integer> Maximum number of consecutive b-frames (now it only enables B GOP structure) Default %d\n", param->bframes);55
+ H0("-b/--bframes <0..16> Maximum number of consecutive b-frames. Default %d\n", param->bframes);56
H1(" --bframe-bias <integer> Bias towards B frame decisions. Default %d\n", param->bFrameBias);57
H0(" --b-adapt <0..2> 0 - none, 1 - fast, 2 - full (trellis) adaptive B frame scheduling. Default %d\n", param->bFrameAdaptive);58
H0(" --[no-]b-pyramid Use B-frames as references. Default %s\n", OPT(param->bBPyramid));59
60
H1(" Format of each line: framenumber frametype QP\n");61
H1(" QP is optional (none lets x265 choose). Frametypes: I,i,K,P,B,b.\n");62
H1(" QPs are restricted by qpmin/qpmax.\n");63
+ H1(" --force-flush <integer> Force the encoder to flush frames. Default %d\n", param->forceFlush);64
+ H1(" 0 - flush the encoder only when all the input pictures are over.\n");65
+ H1(" 1 - flush all the frames even when the input is not over. Slicetype decision may change with this option.\n");66
+ H1(" 2 - flush the slicetype decided frames only.\n");67
H0("\nRate control, Adaptive Quantization:\n");68
H0(" --bitrate <integer> Target bitrate (kbps) for ABR (implied). Default %d\n", param->rc.bitrate);69
H1("-q/--qp <integer> QP for P slices in CQP mode (implied). --ipratio and --pbration determine other slice QPs\n");70
71
H0(" --vbv-maxrate <integer> Max local bitrate (kbit/s). Default %d\n", param->rc.vbvMaxBitrate);72
H0(" --vbv-bufsize <integer> Set size of the VBV buffer (kbit). Default %d\n", param->rc.vbvBufferSize);73
H0(" --vbv-init <float> Initial VBV buffer occupancy (fraction of bufsize or in kbits). Default %.2f\n", param->rc.vbvBufferInit);74
+ H0(" --vbv-end <float> Final VBV buffer emptiness (fraction of bufsize or in kbits). Default 0 (disabled)\n");75
+ H0(" --vbv-end-fr-adj <float> Frame from which qp has to be adjusted to achieve final decode buffer emptiness. Default 0\n");76
H0(" --pass Multi pass rate control.\n"77
" - 1 : First pass, creates stats file\n"78
" - 2 : Last pass, does not overwrite stats file\n"79
80
H0(" --analysis-reuse-mode <string|int> save - Dump analysis info into file, load - Load analysis buffers from the file. Default %d\n", param->analysisReuseMode);81
H0(" --analysis-reuse-file <filename> Specify file name used for either dumping or reading analysis data. Deault x265_analysis.dat\n");82
H0(" --analysis-reuse-level <1..10> Level of analysis reuse indicates amount of info stored/reused in save/load mode, 1:least..10:most. Default %d\n", param->analysisReuseLevel);83
+ H0(" --refine-mv-type <string> Reuse MV information received through API call. Supported option is avc. Default disabled - %d\n", param->bMVType);84
H0(" --scale-factor <int> Specify factor by which input video is scaled down for analysis save mode. Default %d\n", param->scaleFactor);85
- H0(" --refine-intra <int> Enable intra refinement for load mode. Default %d\n", param->intraRefine);86
- H0(" --[no-]refine-inter Enable inter refinement for load mode. Default %s\n", OPT(param->interRefine));87
+ H0(" --refine-intra <0..3> Enable intra refinement for encode that uses analysis-reuse-mode=load.\n"88
+ " - 0 : Forces both mode and depth from the save encode.\n"89
+ " - 1 : Functionality of (0) + evaluate all intra modes at min-cu-size's depth when current depth is one smaller than min-cu-size's depth.\n"90
+ " - 2 : Functionality of (1) + irrespective of size evaluate all angular modes when the save encode decides the best mode as angular.\n"91
+ " - 3 : Functionality of (1) + irrespective of size evaluate all intra modes.\n"92
+ " Default:%d\n", param->intraRefine);93
+ H0(" --refine-inter <0..3> Enable inter refinement for encode that uses analysis-reuse-mode=load.\n"94
+ " - 0 : Forces both mode and depth from the save encode.\n"95
+ " - 1 : Functionality of (0) + evaluate all inter modes at min-cu-size's depth when current depth is one smaller than\n"96
+ " min-cu-size's depth. When save encode decides the current block as skip(for all sizes) evaluate skip/merge.\n"97
+ " - 2 : Functionality of (1) + irrespective of size restrict the modes evaluated when specific modes are decided as the best mode by the save encode.\n"98
+ " - 3 : Functionality of (1) + irrespective of size evaluate all inter modes.\n"99
+ " Default:%d\n", param->interRefine);100
H0(" --[no-]refine-mv Enable mv refinement for load mode. Default %s\n", OPT(param->mvRefine));101
H0(" --aq-mode <integer> Mode for Adaptive Quantization - 0:none 1:uniform AQ 2:auto variance 3:auto variance with bias to dark scenes. Default %d\n", param->rc.aqMode);102
H0(" --aq-strength <float> Reduces blocking and blurring in flat and textured areas (0 to 3.0). Default %.2f\n", param->rc.aqStrength);103
104
H1(" --overscan <string> Specify whether it is appropriate for decoder to show cropped region: undef, show or crop. Default undef\n");105
H0(" --videoformat <string> Specify video format from undef, component, pal, ntsc, secam, mac. Default undef\n");106
H0(" --range <string> Specify black level and range of luma and chroma signals as full or limited Default limited\n");107
- H0(" --colorprim <string> Specify color primaries from undef, bt709, bt470m, bt470bg, smpte170m,\n");108
- H0(" smpte240m, film, bt2020. Default undef\n");109
- H0(" --transfer <string> Specify transfer characteristics from undef, bt709, bt470m, bt470bg, smpte170m,\n");110
+ H0(" --colorprim <string> Specify color primaries from bt709, unknown, reserved, bt470m, bt470bg, smpte170m,\n");111
+ H0(" smpte240m, film, bt2020, smpte428, smpte431, smpte432. Default undef\n");112
+ H0(" --transfer <string> Specify transfer characteristics from bt709, unknown, reserved, bt470m, bt470bg, smpte170m,\n");113
H0(" smpte240m, linear, log100, log316, iec61966-2-4, bt1361e, iec61966-2-1,\n");114
- H0(" bt2020-10, bt2020-12, smpte-st-2084, smpte-st-428, arib-std-b67. Default undef\n");115
+ H0(" bt2020-10, bt2020-12, smpte2084, smpte428, arib-std-b67. Default undef\n");116
H1(" --colormatrix <string> Specify color matrix setting from undef, bt709, fcc, bt470bg, smpte170m,\n");117
- H1(" smpte240m, GBR, YCgCo, bt2020nc, bt2020c. Default undef\n");118
+ H1(" smpte240m, GBR, YCgCo, bt2020nc, bt2020c, smpte2085, chroma-derived-nc, chroma-derived-c, ictcp. Default undef\n");119
H1(" --chromaloc <integer> Specify chroma sample location (0 to 5). Default of %d\n", param->vui.chromaSampleLocTypeTopField);120
H0(" --master-display <string> SMPTE ST 2086 master display color volume info SEI (HDR)\n");121
H0(" format: G(x,y)B(x,y)R(x,y)WP(x,y)L(max,min)\n");122
123
H1("-r/--recon <filename> Reconstructed raw image YUV or Y4M output file name\n");124
H1(" --recon-depth <integer> Bit-depth of reconstructed raw image file. Defaults to input bit depth, or 8 if Y4M\n");125
H1(" --recon-y4m-exec <string> pipe reconstructed frames to Y4M viewer, ex:\"ffplay -i pipe:0 -autoexit\"\n");126
+ H0(" --lowpass-dct Use low-pass subband dct approximation. Default %s\n", OPT(param->bLowPassDct));127
H1("\nExecutable return codes:\n");128
H1(" 0 - encode successful\n");129
H1(" 1 - unable to parse command line\n");130
Refresh
No build results available
Refresh
No rpmlint results available
Login required, please
login
or
signup
in order to comment
Request History
enzokiel created request over 7 years ago
Update to version 2.6
enzokiel accepted request over 7 years ago